JP2006048183A - Data analyzing device and data analyzing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a data analyzing device for intuitively evaluating an analysis result. <P>SOLUTION: This data analyzing device is provided with a data classifying part 4 for classifying a basic data group into a first data group and a second data group based on the value of output attributes, an input attribute condition deciding part 100 for deciding input attribute conditions for dividing the basic data group for gathering the first data group and the second data group, an input attribute condition list graph preparing part 20 for outputting an input attribute condition list graph showing the list of a plurality of decided input attribute conditions and a review graph preparing part 21 for outputting a review graph visually expressing the corresponding relation of at least one of two data groups to be acquired by the division of the basic data group based on the input attribute conditions and at least either the first data group or the second data group to be acquired by classification based on the values of the output attributes in response to an operation on the display screen of the input attribute condition list graph. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an output attribute (object attribute) to be analyzed, such as characteristics of a product manufactured in a manufacturing process, and an input attribute (description attribute) that is an attribute affecting the output attribute, such as a manufacturing process condition. The present invention relates to a data analysis apparatus and a data analysis program for analyzing a cause-and-effect relationship and outputting information indicating the cause-and-effect relationship.

  As an effective method for analyzing the causal relationship between the output attribute and the input attribute, a regression tree method is known. This method creates a tree structure in which the values of the output attributes are well organized at the leaf portions that are sequentially separated by the values of the input attributes.

  FIG. 20 is an example of the regression tree described in FIG. 20 of Patent Document 1, in which the feature value of the threshold voltage VT_N2 of the transistor is an output attribute (object variable), and the device used in each process of manufacturing the transistor Is an analysis result using as an input attribute (an explanatory variable).

  21 and 22 show the review results based on the first branch condition (the branch of Field_Ox = PM1, PM3 and Field_Ox = PM2) of the regression tree (FIG. 20), and the values of the threshold voltage VT_N2 are respectively shown. It is the graph which expressed distribution with the box mustache graph and the histogram.

With reference to FIG. 21 and FIG. 22, it can be confirmed that there is a significant difference in the distribution of the value of the threshold voltage VT_N2 between the PM1 and PM3 use devices and the PM2 use device in the Field_Ox process.
JP 2003-142361 A (publication date: May 16, 2003) Atsushi Otaki, Yuji Horie, Dan Steinberg, "Applied binary tree analysis method-by CART-", Nikka Giren, July 6, 1998, P44-P47

  Conventional regression tree analysis (FIGS. 20 to 22) has the following two problems.

(1) Problem that it is difficult to identify the cause of failure The regression tree in FIG. 20 only shows the condition of the input attribute (explanatory variable) so that the values of the output attribute (object variable) are gathered. No information is shown on which of the two nodes is better and which is worse. For this reason, it is difficult to determine which condition should be specified as the failure factor in the analysis of the failure factor, particularly when the hierarchical structure becomes deep. In addition, it is difficult to prioritize a plurality of conditions appearing on various levels.

  For example, in FIG. 20, if the problem is that the feature value of the threshold voltage VT_N2 (shown next to each node No. in the figure) is large, the feature value in the first branch is Small node No. 1 under the three layers of “Field_Ox = PM1, PM3 (feature amount: 9.498038E-02)”, node No. 10 “first cleaning process_device = # 76 (feature amount: 0.17382)” is not known whether it should be regarded as a problem.

  Further, assuming that this node No. 10 “first cleaning process_apparatus = # 76 (feature amount: 0.17382)” is a problem, the first branch node No. 12 “Field_Ox = PM2 (feature amount: 0.173167)” I don't know which is the bigger problem.

  That is, the conventional regression tree has a first problem that it is difficult to identify the cause of the failure.

(2) Problem that it is difficult to evaluate the analysis result When evaluating the result of the regression tree analysis, the box mustache graph of FIG. 21 and the histogram of FIG. 22 show the point of interest (that is, a clear evaluation index). Therefore, there are problems that it is difficult to make an intuitive judgment and the evaluation of the results varies depending on the user. For example, when referring to the box mustache graph of FIG. 22, the treated product of PM2 has a larger distribution of VT_N2 values than the treated products of PM1 and PM3, but this is regarded as a significant difference or not significant. It depends on the user and is not unique. In the case of this example, a relatively significant difference is seen, so an almost unique judgment may be possible, but in reality, a significant difference does not necessarily appear due to confounding of multiple factors. Absent. That is, the prior art has a second problem that it is difficult to perform unique evaluation because the point of interest when evaluating the analysis result is unclear.

  The present invention has been made in view of the above-described conventional problems, and an object of the present invention is to provide a data analysis apparatus that enables an intuitive evaluation of an analysis result based on a clear index. Another object of the present invention is to provide a data analysis apparatus that clearly identifies a factor of a predetermined output attribute (such as a factor of product characteristic failure) with priority.

  In order to solve the above problems, the data analysis apparatus of the present invention provides the input attribute and the output attribute for a basic data group that is a set of data composed of a plurality of input attributes and output attributes. Is a data analysis device that analyzes the cause-and-effect relationship and outputs the analysis result of the cause-and-effect relationship, and classifies the basic data group into a first data group and a second data group based on the value of the output attribute An input attribute condition that is an input attribute condition for dividing the basic data group into two parts so that the classification means and the first data group and the second data group are respectively combined. Input attribute condition determining means for determining each of the input attribute condition, obtaining a specific input attribute condition from the plurality of input attribute conditions determined by the input attribute condition determining means, and based on the specific input attribute condition At least one of two data groups obtained by bisection of this data group, and at least one of the first data group and the second data group obtained by bisection of the basic data group based on the value of the output attribute And a review graph creating means for outputting a review graph that visually represents the correspondence relationship.

  According to the above configuration, at least one of two data groups obtained by bisection of the basic data group based on a specific input attribute condition and the bisection of the basic data group based on the value of the output attribute A review graph that visually represents the correspondence between at least one of the first data group and the second data group is output. Thereby, the analysis result which shows whether a specific input attribute condition becomes a factor of at least one of the 1st data group and the 2nd data group can be shown to a user. Therefore, the user can intuitively understand the analysis result. In addition, the conventional regression tree problem that “the evaluation of the result varies depending on the user” can be solved.

  These effects are as follows: (1) the classification means classifies the basic data group into the first data group and the second data group based on the value of the output attribute; and (2) the input attribute condition determination means The input attribute condition, which is an input attribute condition for dividing the basic data group into two parts, is set for each of the plurality of input attributes so that the first data group and the second data group are grouped. It depends on what has been decided. That is, the point of interest when evaluating the analysis result is clarified by (1) bisection of the basic data group based on the output attribute and (2) bisection of the basic data group based on the input attribute.

  As described above, according to the above configuration, the object of the present invention of “providing a data analysis apparatus capable of intuitively evaluating an analysis result based on a clear index” can be achieved.

  In addition, according to the above configuration, since the input attribute condition that causes the problem event is extracted for all the input attributes, even if there is a competing factor in the branch condition, it is reliably captured without missing the factor. be able to.

  In order to solve the above problems, the data analysis apparatus of the present invention provides the input attribute and the output attribute for a basic data group that is a set of data composed of a plurality of input attributes and output attributes. Is a data analysis device that analyzes the cause-and-effect relationship and outputs the analysis result of the cause-and-effect relationship, and classifies the basic data group into a first data group and a second data group based on the value of the output attribute An input attribute condition that is an input attribute condition for dividing the basic data group into two parts so that the classification means and the first data group and the second data group are respectively combined. Each of the data constituted by the input attribute condition determining means for determining each of the plurality of input attributes and the output attributes belongs to one of a plurality of groups, Second basic data group creating means for creating a second basic data group representing a data group in which the values of the input attribute and output attribute belonging to the group are averaged in units, and the input attribute condition determining means A specific input attribute condition is acquired from the plurality of input attribute conditions determined in step 1, and at least one of two data groups obtained by bisecting the second basic data group based on the specific input attribute condition And a review graph that visually represents a correspondence relationship between at least one of the two data groups obtained by bisection of the second basic data group based on the condition for classifying the basic data group by the classification unit And a review graph creating means.

  The data analysis apparatus of the present invention preferably further includes an input attribute condition list creation means for outputting a list of the plurality of input attribute conditions determined by the input attribute condition determination means in the data analysis apparatus having the above configuration, The review graph creating means includes a selection operation in which a user selects a specific input attribute condition from the plurality of input attribute conditions included in the list output by the input attribute condition list creating means, and a review graph drawing In response to the instruction, the selected input attribute condition is acquired as the specific input attribute condition, and the review graph is output.

  According to the above configuration, the apparatus further includes an input attribute condition list creating unit that outputs a list of the plurality of input attribute conditions determined by the input attribute condition determining unit, and the review graph creating unit is configured such that the user In response to a selection operation for selecting a specific input attribute condition from a plurality of the input attribute conditions included in the list output by the condition list creation means, the selected input attribute condition is changed to the specific input attribute condition. And the review graph is output. Thereby, the user can obtain a review graph based on an arbitrary input attribute condition among the plurality of input attribute conditions determined by the input attribute condition determining means.

  Further, according to the above configuration, the review graph is output in response to a user giving a review graph drawing instruction. Thereby, a review graph can be output in a timely manner according to a user's request, and a plurality of types of review graphs can be output.

  The specific input attribute condition may be acquired from a selection unit in the data analysis device instead of being acquired by a user selection operation as in the configuration described above. As the selection means, a selection of selecting an input attribute condition having a high index value by using “the ratio of the second data group included in the data group corresponding to the input attribute condition” or the improvement degree of the Gini index method as an index. Or a selection means for selecting all input attribute conditions included in the list output by the input attribute condition list creation means.

  Further, as the reference (index) for determining the input attribute condition in the input attribute condition determining means, the improvement level of the Gini index method, the frequency difference value, a combination thereof, and the like can be used. In the first data group, the frequency difference value is calculated as the first frequency for all the numerical values of each input attribute as the first frequency, and in the second data group, A frequency calculation step for calculating a ratio of data whose input attribute is less than or equal to the numerical value as a second frequency, and a difference for calculating a difference value between the first frequency and the second frequency for all the numerical values of each input attribute It can be calculated by calculating steps.

  In the data analysis device of the present invention, preferably, in the data analysis device configured as described above, the input attribute condition determined by the input attribute condition determining means is “the second data group in the data group that satisfies the input attribute condition”. And the review graph creating means includes at least a data group that satisfies the input attribute condition, and a data group that does not satisfy the input attribute condition. This is a configuration for outputting a review graph that visually represents a correspondence relationship with a data group that satisfies the output attribute condition corresponding to the second data group.

  According to the above configuration, at least a review graph that visually represents a correspondence relationship between a data group that satisfies a specific input attribute condition and the second data group is output. Thereby, it is possible to present to the user an analysis result indicating whether the specific input attribute condition is a factor of the output attribute corresponding to the second data group. Therefore, the user can intuitively understand the analysis result.

  The data analysis apparatus of the present invention is preferably the data analysis apparatus configured as described above, wherein the list of input attribute conditions includes the basic attribute relating to each of the plurality of input attribute conditions determined by the input attribute condition determination means. A second data group corresponding number representing the number of data corresponding to the second data group that satisfies the input attribute condition in the data group, and data satisfying the input attribute condition in the basic data group The second data group separation degree indicating the ratio of the number of data in which the second data group is included is configured to be shown together with the input attribute condition.

  According to the above configuration, the list of the input attribute conditions includes the second data group corresponding number representing the number of data satisfying the input attribute condition in the basic data group and corresponding to the second data group; The second data group separation degree representing the ratio of the number of data in which the second data group is included in the data satisfying the input attribute condition in the basic data group is indicated together with the input attribute condition. Yes. For this reason, the user refers to the second data group hit count and the second data group separation degree shown together with the list of input attribute conditions, so that each input attribute condition corresponds to the second data group. You can know how much it contributes to the generation of attributes. Therefore, it is possible to create a review graph with a high possibility that the user can find the factor of the output attribute corresponding to the second data group.

  Further, according to the above configuration, both the number of corresponding second data groups and the second data group separation degree are shown in the list of input attribute conditions. Accordingly, the user can select one of (1) the number of corresponding second data groups, (2) the degree of second data group separation, (3) the number of corresponding second data groups and the degree of second data group separation depending on the application. It can be selected as an index and a review graph can be created based on the index. Therefore, the user can easily create a review graph according to the application.

  The data analysis device of the present invention is preferably the data analysis device configured as described above, wherein the input attribute condition is determined in the basic data group for the plurality of input attribute conditions determined by the input attribute condition determination means. A second data group corresponding number that represents the number of data satisfying and corresponding to the second data group, and corresponding to the second data group in the data that satisfies the input attribute condition in the basic data group Rank determining means for performing ranking according to importance as a factor of an output attribute condition corresponding to the second data group, using at least one of the second data group separation degrees indicating the ratio of the number of data to be performed as an index; The input attribute condition list creating means is configured to show the ranking result by the rank determining means in the list of input attribute conditions.

  According to the above configuration, the input attribute condition list creation unit also shows the input attribute that also indicates the result of prioritization based on the second data group hit count and / or the second data group separation degree determined by the rank determination unit Create a list of conditions. By referring to this list of input attribute conditions, the user can grasp and understand a plurality of input attribute conditions that cause the output attribute corresponding to the second data group, and know the order of importance. it can.

  Therefore, according to the above configuration, another object of the present invention is to “provide a data analysis apparatus that clearly specifies a factor of a predetermined output attribute (such as a factor of product characteristic failure) with priority ranking”. Can be achieved.

  Further, according to the above configuration, the user performs a selection operation while referring to the order (priority order) of the input attribute conditions in the list of input attribute conditions (that is, while navigating the provisional order for reviewing the analysis results). A review graph can be created by performing.

  The data analysis apparatus of the present invention is the data analysis apparatus having the above configuration, wherein each of the data composed of a plurality of input attributes and output attributes in the basic data group or the second basic data group includes: Each is provided with a discrimination key, and the review graph creating means sets the target data group to be the target of the review graph as the basic data group or the second basic data group, and for the target data group Both the first graph showing the change of the input attribute corresponding to the acquired specific input attribute condition for each of the discrimination keys and the second graph showing the change of the output attribute for each of the discrimination keys A graph to be output as the review graph. In the first graph, the output attribute condition that satisfies the input attribute condition and corresponds to the second data group is output. May be configured to marking process is performed on data that satisfies.

  In the above configuration, the determination key may be assigned a different determination key for each piece of data in the basic data group, or may be assigned a different determination key for each group of data in the basic data group.

  The data analysis apparatus according to the present invention is the data analysis apparatus having the above-described configuration, wherein the basic data group includes a plurality of data corresponding to different time points, and the target data group to be created as the review graph is the basic data. A first trend graph showing the time change of the input attribute corresponding to the acquired specific input attribute condition for the target data group and the time change of the output attribute A first review graph showing both the second trend graph and the second trend graph are output as the review graph. In the first trend graph, the input attribute condition is satisfied, and the second data group includes A configuration in which marking processing is performed on data corresponding to the corresponding output attribute condition may be employed.

  According to the above configuration, the specific data acquired from the condition that “the second data group is grouped into a data group that satisfies the input attribute condition and the first data group is grouped into a data group that does not satisfy the input attribute condition”. The first review graph including the first trend graph indicating the time change of the input attribute corresponding to the input attribute condition is output. Further, according to the above configuration, in the first trend graph, marking processing is performed on data satisfying the input attribute condition and satisfying the output attribute condition corresponding to the second data group. Accordingly, the user can easily know from the first trend graph the period during which the specific input attribute condition is the cause of the output attribute corresponding to the second data group. Further, according to the above configuration, the first review graph includes the first trend graph and the second trend graph indicating the time change of the output attribute. By comparing, it is possible to immediately understand the synchronization (relevance) of the time change of the specific input attribute condition with the time change of the output attribute.

  Note that the “marking process” is a process for marking a part indicating specific data in a review graph so that it can be distinguished from a part indicating other data, for example, other data in a part indicating specific data. Processing that gives a color different from the part that indicates, Processing that changes the shape of the part that shows specific data to a shape that is different from the part that shows other data, Process that puts a mark (for example, a surrounding line) only on the part showing specific data Etc.

  The data analysis apparatus according to the present invention is the data analysis apparatus having the above configuration, wherein the basic data group includes a plurality of data corresponding to different positions in a plane, and the review graph creation means is a target for creating the review graph. The target data group is the basic data group or the second basic data group, and the first in-plane distribution of the input attribute corresponding to the acquired specific input attribute condition is expressed for the target data group. And a second review graph showing both the second map representing the in-plane distribution of the output attribute as the review graph, and the first map satisfies the input attribute condition. And the structure by which the marking process is given to the data which satisfy | fills the output attribute conditions corresponding to a said 2nd data group may be sufficient.

  According to the above configuration, the specific data acquired from the condition that “the second data group is grouped into a data group that satisfies the input attribute condition and the first data group is grouped into a data group that does not satisfy the input attribute condition”. The second review graph including the first map indicating the in-plane distribution of the input attribute corresponding to the input attribute condition is output. Further, according to the above configuration, in the first map, marking processing is performed on data satisfying the input attribute condition and satisfying the output attribute condition corresponding to the second data group. As a result, the user can easily know from the first map the in-plane position where the specific input attribute condition may cause the output attribute corresponding to the second data group. Furthermore, according to the above configuration, since the second review graph shows the second map indicating the in-plane distribution of the output attribute together with the first map, the user can display these in-plane distributions. By comparing, it is possible to immediately understand the synchronism (relationship) between the in-plane distribution of the specific input attribute condition and the in-plane distribution of the output attribute.

  The data analysis apparatus according to the present invention is the data analysis apparatus having the above-described configuration, wherein the review graph creating means sets the target data group to be created as the review graph as the basic data group or the second basic data group. A third review graph, which is a scatter diagram showing the relationship between the input attribute corresponding to the acquired specific input attribute condition and the output attribute, is output as the review graph for the target data group. Yes, the scatter diagram may be configured such that marking processing is performed on data that satisfies the input attribute condition and satisfies the output attribute condition corresponding to the second data group.

  According to the said structure, the 3rd review graph which is a scatter diagram which shows the relationship between the input attribute corresponding to the acquired said specific input attribute condition, and the said output attribute can be output. By referring to this scatter diagram, the user can satisfy the condition of “the second data group is grouped into a data group that satisfies the input attribute condition and the first data group is grouped into a data group that does not satisfy the input attribute condition”. It is possible to immediately understand the correlation between the specific input attribute condition acquired from the inside and the output attribute viewed in multiple values.

  In the data analysis apparatus according to the present invention, in the data analysis apparatus having the above-described configuration, the review graph creating means instructs the grouping by another input attribute condition different from the input attribute condition corresponding to the axis of the scatter diagram. In response to this, the configuration may be such that the data satisfying the other input attribute condition and the data not satisfying the other input attribute condition are shown in a scatter diagram in different forms.

  According to the above configuration, the user refers to the scatter diagram to obtain the correlation between the specific input attribute condition acquired from the input attribute condition determined by the input attribute condition determining unit and the output attribute, The correlation between the input attribute condition and the output attribute can be understood independently.

  The data analysis apparatus according to the present invention is the data analysis apparatus having the above-described configuration, wherein the review graph creating means sets the target data group to be created as the review graph as the basic data group or the second basic data group. The target data group is separated into a data group satisfying the acquired specific input attribute condition and a data group not satisfying the input attribute condition, and the frequency distribution of the output attribute in each of the data groups is shown. The structure which outputs a 4th review graph as said review graph may be sufficient.

  According to the above configuration, the basic data group is separated into a data group that satisfies the acquired specific input attribute condition and a data group that does not satisfy the input attribute condition, and the output in each of the data groups A fourth review graph indicating the frequency distribution of attributes can be output as the review graph. By referring to the fourth review graph indicating this frequency distribution, the user can obtain the specific input attribute condition acquired from the input attribute conditions determined by the input attribute condition determination means and the exclusive condition, The degree of separation between the first data group and the second data group can be intuitively understood.

  In order to solve the above problems, the data analysis apparatus of the present invention provides the input attribute and the output attribute for a basic data group that is a set of data composed of a plurality of input attributes and output attributes. Analyzing the causal relationship and outputting the analysis result of the causal relationship, and classifying the basic data group into a first data group and a second data group based on the value of the output attribute A plurality of input attributes, the input attribute condition being an input attribute condition for bisecting the basic data group so that the first data group and the second data group are each grouped Each of the input attribute condition determining means to determine, and two or more specific input attribute conditions are acquired from the plurality of input attribute conditions determined by the input attribute condition determining means, and the specified 2 or more of The first data group and the second data group obtained by bifurcation of a data group extracted from the basic data group based on a condition entangled with an input attribute condition, and the basic data group based on a value of the output attribute And a composite factor graph creating means for outputting a composite factor graph representing a correspondence relationship between at least one of the above.

  According to the above configuration, the data group extracted from the basic data group based on a condition in which two or more specific input attribute conditions are entangled and the basic data group based on a value of the output attribute are obtained by bisection. A composite factor graph that visually represents a correspondence relationship with at least one of the first data group and the second data group is output. As a result, it is possible to present to the user an analysis result indicating whether a condition in which two or more specific input attribute conditions are entangled is a factor of at least one of the first data group and the second data group. Therefore, the user can intuitively understand the analysis result. In addition, the conventional regression tree problem that “the evaluation of the result varies depending on the user” can be solved.

  These effects are as follows: (1) the classification means classifies the basic data group into the first data group and the second data group based on the value of the output attribute; and (2) the input attribute condition determination means The input attribute condition, which is an input attribute condition for dividing the basic data group into two parts, is set for each of the plurality of input attributes so that the first data group and the second data group are grouped. It depends on what has been decided. That is, the point of interest when evaluating the analysis result is clarified by (1) bisection of the basic data group based on the output attribute and (2) bisection of the basic data group based on the input attribute.

  As described above, according to the above configuration, the object of the present invention of “providing a data analysis apparatus capable of intuitively evaluating an analysis result based on a clear index” can be achieved.

  In addition, according to the above configuration, since the input attribute condition that causes the problem event is extracted for all the input attributes, even if there is a competing factor in the branch condition, it is reliably captured without missing the factor. be able to.

  The invention relating to the data analysis apparatus including the review graph creating means and the invention relating to the data analysis apparatus including the composite factor graph creating means have the same technical features as the classification means and the input attribute condition determining means, Special technical feature (review graph creation means and composite factor graph creation means) that clearly shows the contribution of the invention to the prior art, "Providing a data analysis device that enables intuitive evaluation of analysis results based on various indicators" ) To form a single general inventive concept.

  Here, “a condition in which two or more input attribute conditions are entangled” means a condition obtained by combining two or more input attribute conditions, for example, a logical product (two or more input attribute conditions). A condition that satisfies all of the input attribute conditions) a negative logical product of two or more input attribute conditions (a condition that satisfies all the exclusive conditions for each of the two input attribute conditions), and an exclusive OR of the two input attribute conditions ( A condition that satisfies only one of the two input attribute conditions).

  The data analysis apparatus of the present invention preferably further includes an input attribute condition list creation means for outputting a list of the plurality of input attribute conditions determined by the input attribute condition determination means in the data analysis apparatus having the above configuration, The composite factor graph creating means includes a selection operation in which a user selects a specific input attribute condition from a plurality of the input attribute conditions included in the list output by the input attribute condition list creating means, and a composite factor In response to performing a graph drawing instruction, the input attribute condition including the selected input attribute condition is acquired as the two or more specific input attribute conditions, and the composite factor graph is output. It is a configuration.

  According to the above configuration, the apparatus further includes an input attribute condition list creating unit that outputs a list of the plurality of input attribute conditions determined by the input attribute condition determining unit, wherein the composite factor graph creating unit is configured so that the user can input the input attribute condition list. In response to a selection operation for selecting a specific input attribute condition from a plurality of the input attribute conditions included in the list output by the attribute condition list creating means, an input attribute condition including the selected input attribute condition is The two or more specific input attribute conditions are acquired, and the composite factor graph is output. Thereby, the user can obtain a composite factor graph based on a combination of arbitrary input attribute conditions among the plurality of input attribute conditions determined by the input attribute condition determining means.

  Further, according to the above configuration, the composite factor graph is output in response to a user giving a composite factor graph drawing instruction. As a result, a composite factor graph can be output in a timely manner according to the user's wishes, and a plurality of types of composite factor graphs can be output.

  Note that the two or more specific input attribute conditions may be acquired from a selection unit in the data analysis apparatus instead of being acquired by a user selection operation as in the above configuration. As the selection means, a selection of selecting an input attribute condition having a high index value by using “the ratio of the second data group included in the data group corresponding to the input attribute condition” or the improvement degree of the Gini index method as an index. Or a selection means for selecting all combinations of a predetermined number of input attribute conditions from among the input attribute conditions included in the list output by the input attribute condition list creation means. Further, as the reference (index) for determining the input attribute condition in the input attribute condition determining means, the improvement degree of the Gini index method, the above-described frequency difference value, a combination thereof, and the like can be used.

  In the data analysis device of the present invention, preferably, in the data analysis device configured as described above, the input attribute condition determined by the input attribute condition determining means is “the second data group in the data group that satisfies the input attribute condition”. The first data group is grouped into a data group that does not satisfy the input attribute condition ”, and the input attribute condition list includes a plurality of items determined by the input attribute condition determining unit. A second data group corresponding number that represents the number of data that satisfies the input attribute condition in the basic data group and that corresponds to the second data group, and the basic data group And the second data group separation degree indicating the ratio of the number of data in which the second data group is included in the data satisfying the input attribute condition is also shown together with the input attribute condition It is a configuration.

  According to the above configuration, the list of the input attribute conditions includes the second data group corresponding number representing the number of data satisfying the input attribute condition in the basic data group and corresponding to the second data group; The second data group separation degree representing the ratio of the number of data in which the second data group is included in the data satisfying the input attribute condition in the basic data group is indicated together with the input attribute condition. Yes. For this reason, the user refers to the second data group hit count and the second data group separation degree shown together with the list of input attribute conditions, so that each input attribute condition corresponds to the second data group. You can know how much it contributes to the generation of attributes. Therefore, it is possible to create a composite factor graph with a high possibility that the user can find the factor of the output attribute corresponding to the second data group.

  Further, according to the above configuration, both the number of corresponding second data groups and the second data group separation degree are shown in the list of input attribute conditions. Accordingly, the user can select one of (1) the number of corresponding second data groups, (2) the degree of second data group separation, (3) the number of corresponding second data groups and the degree of second data group separation depending on the application. It can be selected as an index, and a composite factor graph can be created based on the index. Therefore, the user can easily create a composite factor graph according to the application.

  In the data analysis device of the present invention, preferably, in the data analysis device configured as described above, the input attribute condition determined by the input attribute condition determining means is “the second data group in the data group that satisfies the input attribute condition”. Is a condition determined such that the first data group is collected in a data group that does not satisfy the input attribute condition, and for the plurality of input attribute conditions determined by the input attribute condition determining means, A second data group corresponding number representing the number of data corresponding to the second data group that satisfies the input attribute condition in the basic data group, and data satisfying the input attribute condition in the basic data group Output corresponding to the second data group, using as an index at least one of the second data group separation degrees indicating the ratio of the number of data corresponding to the second data group The input attribute condition list creation means further includes a ranking determination means for ranking by importance as a factor of the sex condition, and the input attribute condition list creation means shows a result of ranking by the ranking determination means in the list of input attribute conditions. .

  According to the above configuration, the input attribute condition list creation unit also shows the input attribute that also indicates the result of prioritization based on the second data group hit count and / or the second data group separation degree determined by the rank determination unit Create a list of conditions. By referring to this list of input attribute conditions, the user can grasp and understand a plurality of input attribute conditions that cause the output attribute corresponding to the second data group, and know the order of importance. it can.

  Therefore, according to the above configuration, another object of the present invention is to “provide a data analysis apparatus that clearly specifies a factor of a predetermined output attribute (such as a factor of product characteristic failure) with priority ranking”. Can be achieved.

Further, according to the above configuration, the user performs a selection operation while referring to the order (priority order) of the input attribute conditions in the list of input attribute conditions (that is, while navigating the provisional order for performing the complex factor analysis). By performing the above, a composite factor graph can be created.
In the data analysis apparatus according to the present invention, the input attribute condition determined by the input attribute condition determination unit in the data analysis apparatus having the above configuration is “the second data group is grouped into a data group that satisfies the input attribute condition, The composite factor graph creating means includes a plurality of the input attributes determined by the input attribute condition determining means, such that the first data group is collected into a data group that does not satisfy the input attribute condition. Result of obtaining two specific input attribute conditions from the conditions, and calculating the number of data satisfying both of the two input attribute conditions in the basic data group and corresponding to the second data group In the basic data group, the number of data satisfying only one of the two input attribute conditions and corresponding to the second data group is calculated. Fruit may be configured and outputs a first composite factor graph shown respectively as the multiple factors graph.

  According to the above configuration, for the two specific input attribute conditions acquired from the plurality of input attribute conditions determined by the input attribute condition determining means, Both the input attribute conditions are satisfied and the number of data corresponding to the second data group is calculated, and in the basic data group, only one of the two input attribute conditions is satisfied, In addition, it is possible to output, as the composite factor graph, the first composite factor graph in which the results of calculating the number of data corresponding to the second data group are respectively shown. Thereby, by referring to the first composite factor graph, the user can determine whether the other input attribute condition of the two specific input attribute conditions is one input attribute condition of the two specific input attribute conditions. It is possible to know whether it is a complex factor (or a factor that has a master-slave relationship) or a factor that is independent of the one input attribute condition. Therefore, a plurality of factors (or a plurality of factors having a master-slave relationship) that are complex factors and a plurality of factors that are independent from each other can be searched from the input attribute conditions determined by the input attribute condition determining means. .

  In the data analysis apparatus according to the present invention, the input attribute condition determined by the input attribute condition determination unit in the data analysis apparatus having the above configuration is “the second data group is grouped into a data group that satisfies the input attribute condition, The composite factor graph creating means includes a plurality of the input attributes determined by the input attribute condition determining means, such that the first data group is collected into a data group that does not satisfy the input attribute condition. Two or more specific input attribute conditions are acquired from the conditions by ranking, and the input attribute conditions are satisfied in the basic data group in the order of the ranking of the input attribute conditions acquired by ranking. In addition, a second composite factor graph showing the result of sequentially integrating the second data group corresponding number representing the number of data corresponding to the second data group is output as the composite factor graph. It may be in a configuration of.

  According to the above configuration, the number of pieces of data satisfying the input attribute condition in the basic data group and corresponding to the second data group is represented in the order of the rank of the input attribute condition acquired with the ranking. A second composite factor graph showing the result of sequentially integrating the number of hits in the second data group can be output as the composite factor graph. By referring to the second composite factor graph, the user can substantially determine the validity of the two or more acquired input attribute conditions, that is, the input attribute conditions that cause the output attribute corresponding to the second data group. It can be confirmed whether it is possible to select without omission.

  In the data analysis apparatus according to the present invention, the input attribute condition determined by the input attribute condition determination unit in the data analysis apparatus having the above configuration is “the second data group is grouped into a data group that satisfies the input attribute condition, The composite factor graph creating means is a condition determined such that the first data group is collected in a data group that does not satisfy the input attribute condition. A data group satisfying all of the input attribute conditions and a data group satisfying all of the exclusive conditions for each of the two or more acquired input attribute conditions are respectively extracted, and the frequency distribution of the output attribute in each of the data groups The third composite factor graph indicating the above may be output as the composite factor graph.

  According to the above configuration, the output attribute in each of the data group that satisfies all of the plurality of acquired input attribute conditions and the data group that satisfies all of the exclusive conditions for each of the acquired plurality of input attribute conditions It is possible to output a third composite factor graph showing the frequency distribution. Thereby, by referring to the third composite factor graph, the user can confirm the mutual influence of the plurality of input attribute conditions of interest.

  In the data analysis apparatus of the present invention, the input attribute condition determination unit includes an analysis data group extraction unit that extracts an analysis data group to be analyzed from the basic data group after classification by the classification unit; An input attribute condition, which is an input attribute condition for bisecting the analysis data group, is set for each of the plurality of input attributes so that the first data group and the second data group are respectively combined. The first input attribute condition determining means for determining each of the input attribute conditions, and the input attribute condition having the highest probability of cutting out the second data group among the input attribute conditions determined by the first input attribute condition determining means. Branch condition determining means for determining the analysis data group into two data groups based on the branch condition determined by the branch condition determining means, A second input attribute condition for selecting a highly important input attribute condition as a factor of the output attribute condition corresponding to the second data group from the input attribute conditions determined by the first input attribute condition determining means And the analysis data group extraction means extracts only a data group having a smaller ratio of the second data group as a new analysis data group out of the two data groups divided by the division means. And a series of processes consisting of the processing by the analysis data group extracting means, the processing by the first input attribute condition determining means, the processing by the branch condition determining means, and the processing by the dividing means are repeatedly executed. The second input attribute condition determining means is determined for each of the input attributes by the repetition process of the first input attribute condition determining means. The input attribute condition having a high degree of importance is selected from a plurality of input attribute conditions in the input attribute, and the selected input attribute condition is determined as an output of the input attribute condition determining means. preferable.

  According to the above configuration, since the second input attribute condition determining means selects the input attribute condition having a high importance based on the importance as the factor of the output attribute condition corresponding to the second data group, Regardless of how complicated the regression tree is, the cause of the output attribute condition corresponding to the second data group can be clearly understood.

  Further, according to the above configuration, the second input attribute condition determining means selects only a highly important condition among a plurality of input attribute conditions in the same input attribute determined for each repetition process. In order to obtain detailed factor analysis results, the above-mentioned iterative process is not complicated, and the output attribute condition factor corresponding to the second data group can be determined with high accuracy while maintaining a very simple form. Can be determined.

  Furthermore, according to the above configuration, the analysis data group extraction is performed only for the data group having the smaller ratio of the second data group among the two data groups (nodes) divided (branched) by the dividing unit. A series of processing consisting of processing by means, processing by the first input attribute condition determining means, processing by the branch condition determining means, and processing by the dividing means is repeated. In this way, since the output attribute factor corresponding to the second data group is determined only for the data group in which the cause of the output attribute corresponding to the second data group is not yet found, the input attribute condition that causes the output attribute corresponding to the second data group is determined. By excluding the influence of the input attribute condition determined as the main factor of the second data group in the previous iterative process, the new factor of the second data group can be extracted with high accuracy and efficiency.

  In order to solve the above problems, the data analysis apparatus of the present invention provides the input attribute and the output attribute for a basic data group that is a set of data composed of a plurality of input attributes and output attributes. A data analysis apparatus for analyzing the causal relationship of the data and extracting information indicating the causal relationship, and classifying the basic data group into a first data group and a second data group based on the value of the output attribute Classification means, analysis data group extraction means for extracting an analysis data group to be analyzed from the basic data group after classification by the classification means, the first data group and the second data group, Are first input attribute condition determining means for determining, for each of the plurality of input attributes, an input attribute condition that is an input attribute condition for dividing the analysis data group into two. And before Among the input attribute conditions determined by the first input attribute condition determining means, the branch condition determining means for determining the input attribute condition having the highest probability of cutting out the second data group as a branch condition; A splitting unit that divides the analysis data group into two based on the branch condition determined by the branching condition determination unit, and the analysis data group extraction unit includes the data group divided by the splitting unit, Only a data group having a smaller ratio of the second data group is extracted as a new analysis data group, processing by the analysis data group extraction means, processing by the first input attribute condition determination means, and by the branch condition determination means A series of processes including processes and processes by the dividing unit are repeatedly executed.

  According to the above configuration, only the data group having the smaller ratio of the second data group among the two data groups (nodes) divided (branched) by the dividing unit is analyzed by the analysis data group extracting unit. A series of processing consisting of processing, processing by the first input attribute condition determining means, processing by the branch condition determining means, and processing by the dividing means is repeated. In this way, since the output attribute factor corresponding to the second data group is determined only for the data group in which the cause of the output attribute corresponding to the second data group is not yet found, the input attribute condition that causes the output attribute corresponding to the second data group is determined. By excluding the influence of the input attribute condition determined as the main factor of the second data group in the previous iterative process, the new factor of the second data group can be extracted with high accuracy and efficiency.

  In addition, according to the above configuration, since the input attribute condition that causes the problem event is extracted for all the input attributes, even if there is a competing factor in the branch condition, it is reliably captured without missing the factor. be able to.

  Note that “the accuracy of cutting out the second data group” refers to the accuracy of “the second data group is collected in the data group that satisfies the input attribute condition”. As the “accuracy of cutting out the second data group”, the improvement degree of the Gini index method, “the ratio of the second data group included in the data group corresponding to the input attribute condition”, the first frequency and the first The ratio with the frequency of 2, the frequency difference value described above, a combination thereof, and the like can be used.

  The data analysis apparatus and the data analysis program according to the present invention provide two data obtained by bifurcation of a basic data group based on a specific input attribute condition selected from a plurality of the input attribute conditions determined by the input attribute condition determining means Since a review graph that visually represents the correspondence between the group and two data groups (first data group and second data group) obtained by bisectioning the basic data group based on the value of the output attribute is output, Intuitive evaluation of analysis results is possible.

  Further, the data analysis apparatus and the data analysis program of the present invention are obtained by bisecting the second basic data group based on a specific input attribute condition selected from the plurality of input attribute conditions determined by the input attribute condition determining means. A review graph that visually represents the correspondence between the two data groups obtained and the two data groups obtained by bisectioning the second basic data group based on the conditions for classifying the basic data group by the classification means Can be intuitively evaluated.

  Further, the data analysis apparatus and the data analysis program of the present invention are based on a condition in which two or more specific input attribute conditions selected from the plurality of input attribute conditions determined by the input attribute condition determining means are entangled. A composite representing a correspondence relationship between a data group extracted from the basic data group and two data groups (first data group and second data group) obtained by dividing the basic data group based on the value of the output attribute Since the factor graph is output, the analysis result can be intuitively evaluated.

  In the data analysis apparatus of the present invention, the classification means for classifying the basic data group into the first data group and the second data group based on the value of the output attribute, the first data group and the second data group, Input attribute condition determining means for determining an input attribute condition for dividing the basic data group into two parts, and an input attribute for outputting a list of a plurality of input attribute conditions determined by the input attribute condition determining means And a condition list creating means. In this list of input attribute conditions, priorities are assigned to a plurality of input attribute conditions based on the number of second data group hits and / or the second data group separation degree. In this case, it is possible to clearly specify the factor of the predetermined output attribute by assigning priority.

  Further, the data analysis apparatus and the data analysis program of the present invention sequentially outputs the output attribute factor corresponding to the second data group to only the data group for which the output attribute factor corresponding to the second data group is not found. Since the input attribute condition is determined, the influence of the input attribute condition determined as the main factor of the second data group in the previous iterative process is excluded, and the new factor of the second data group is highly accurate. And it can extract efficiently.

  One embodiment of the present invention will be described below.

  First, the structure of the data analysis apparatus of this embodiment is demonstrated based on FIG. As shown in FIG. 1, the data analysis apparatus includes a basic data group storage unit 1, a classification condition setting unit 3, a data classification unit (classification means) 4, a post-classification basic data group storage unit 5, an analysis data group extraction unit (analysis). Data group extracting means) 6, first input attribute condition determining section (first input attribute condition determining means) 12, analysis result data storage section 13, branch condition determining section (branch condition determining means) 14, data dividing section ( (Dividing means) 15, end condition determination unit 16, defective product data calculation unit 17, second input attribute condition determination unit (second input attribute condition determination unit) 19, input attribute condition list graph creation unit (input attribute condition list) Creating means) 20, review graph creating section (second basic data group creating means, review graph creating means) 21, composite factor graph creating section (composite factor graph creating means) 22, and rank determining section (rank determining means) 2 It is equipped with a. The analysis data group extraction unit (analysis data group extraction unit) 6, the first input attribute condition determination unit (first input attribute condition determination unit) 12, the analysis result data storage unit 13, the branch condition determination unit ( Branch condition determining means) 14, data dividing section (dividing means) 15, end condition determining section 16, defective product data calculating section 17, and second input attribute condition determining section (second input attribute condition determining means) 19 Constitutes the input attribute condition determining unit 100 as the input attribute condition determining means in the claims.

  Further, the data analysis apparatus creates an input device 30 such as a keyboard and a pointing device for inputting classification condition setting information to the classification condition setting unit 3 of the data analysis apparatus, and an input attribute condition list graph of the data analysis apparatus. An input device 31 such as a keyboard, a pointing device, and a switch for sending a user instruction to the unit 20 and a display device 32 that displays a review graph and the like are connected. Note that the input device 30 and the input device 31 may be realized by a single input device, for example, a keyboard. Instead of the display device 32, a review graph or the like may be output using an output device other than the display device, for example, a printing device.

  Next, the operation of the data analysis apparatus of the present embodiment will be described by taking as an example the case where the data group DA in Table 1 is a basic data group. The basic data group DA in Table 1 is stored in the basic data group storage 1 such as a hard disk.

  The basic data group DA in Table 1 relates to products in which a plurality of elements (functional elements) such as liquid crystal panels and IC chips manufactured in the manufacturing process are arranged on a substrate, and the characteristics of each element for each element. It is a sample of a data group in which data, inspection data, process data, and the like are arranged.

  In Table 1, “element number” is a unique number assigned to each element, “substrate number” is the number of the substrate to which each element belongs, and “element coordinate number” is the number of each element in the substrate. The numbers of the coordinate positions in are respectively represented. As shown in Table 1, each of the plurality of data included in the basic data group DA includes eight input attributes 1 to 8, an output attribute, an element number, a board number, and an element coordinate number. Each substrate is manufactured in the same manufacturing process, and the numbers are given in the order of manufacturing time.

  The “output attribute” represents an attribute to be analyzed such as element characteristic data, and “input attributes 1 to 8” include “manufacturing process conditions and in-line inspection data (inspection results during manufacturing)”. Indicates an attribute that affects the “output attribute”.

  FIG. 2 is a trend graph showing the transition of the “output attribute” value for each board, and shows the average value of the “output attribute” values of each element in the same board for each board. FIG. 3 is a map showing the in-plane distribution of the “output attribute” value. The element coordinate position where the average value of the “output attribute” value of each element in the same “element coordinate number” is large is shown in red (on the drawing). The other coordinate positions are shown in green (indicated by hatching in the drawing).

  Now, an element having an “output attribute” value larger than 2.5 (“output attribute”> 2.5) is regarded as a defective product, and the cause of this failure is extracted from “input attributes 1 to 8”. The operation of the data analysis apparatus in FIG. 1 will be described by taking the case of evaluating the result as an example.

  In the trend graph of FIG. 2, the substrate corresponding to the plot above the broken line A indicates a substrate including many defective products (elements with “output attribute”> 2.5). In the map of FIG. 3, the positions of P1 and P2 indicate coordinate positions including many defective products (elements with “output attribute”> 2.5), and the positions of P3, P4, and P5 are defective products ( A coordinate position with a small number of “output attributes”> 2.5 elements) is shown.

  Hereinafter, the operation of the data analysis apparatus of FIG. 1 will be described with reference to the flowchart shown in FIG. The basic data group DA in Table 1 is stored in the basic data group storage unit 1 such as a hard disk.

[Step 1]
The classification condition setting unit 3 sets “output attribute> 2.5”, which is a condition for defective products, according to predetermined setting information or according to setting information input by the user from the input device 30 such as a keyboard or a mouse. And output to the data classification unit 4 (S1).

[Step 2]
Next, the data classification unit 4 converts the basic data group DA into the first data group based on the value of the output attribute by the following logic (1) (2) based on the classification condition output from the classification condition setting unit 3. The data is classified into DA1 and second data group DA2, and as shown in Table 2, a classification flag corresponding to each data group is assigned (S2).

  Hereinafter, the data group in Table 2 is referred to as a post-classification basic data group DA00. The post-classification basic data group DA00 is stored in the post-classification basic data group storage 5 such as a hard disk.

(1) “Output attribute ≦ 2.5” → DA1
(2) “Output attribute> 2.5” → DA2
Here, the second data group DA2 is a defective product data group (“output attribute> 2.5”), and the first data group DA1 is a non-defective product data group (“output attribute ≦ 2.5”). . Hereinafter, the first data group DA1 is appropriately referred to as a non-defective product (OK product) data group, and the second data group DA2 is appropriately referred to as a defective product (NG product) data group.

  The classification by the data classification unit 4 is not limited to the above logic, and may be performed according to the logic shown in the following classification condition based on whether or not it corresponds to specific character data such as a determination result of defective products. .

(1 ') “Output attribute ≠“ defective product ”” → DA1
(2 ′) “Output attribute =“ defective product ”” → DA2
Further, based on a logical sum or logical product of a plurality of conditions expressed as a comparison result with a plurality of thresholds, etc., the classification may be performed according to the logic shown in the following classification conditions.

(1 “) yth1 <output attribute ≦ yth2 → DA1 (yth1, yth2: threshold)
(2 “) Output attribute ≦ yth1 OR Output attribute> yth2 → DA2
Furthermore, the following classification conditions based on the logical sum or logical product of the conditions expressed as a comparison result with each threshold value for each output attribute, such as multiple output attributes (output attribute 1, output attribute 2) You may classify | categorize according to logic as shown in.

(1 ″ ′) Output attribute 1 ≦ yth1 OR Output attribute 2> yth2 → DA1
(2 ″ ′) Output attribute 1> yth1 AND Output attribute 2 ≦ yth2 → DA2
[Step 3]
The analysis data group extraction unit 6 extracts an analysis data group DA00 ′ to be analyzed from the classified basic data group DA00 and sends it to the first input attribute condition determination unit 12.

  In this first process, the same data as the post-classification basic data group DA00 is extracted as the analysis data group DA00 ′. However, in the process of repeated processing described later, another data group output by the data dividing unit 15 is extracted. The

[Step 4]
The first input attribute condition determination unit 12 determines an input attribute condition for each of the input attributes 1 to 8. The input attribute condition is such that the first data group DA1 of the non-defective product and the second data group DA2 of the defective product are grouped together, that is, one of the two data groups obtained by the bisection is set to the first data group DA1. This is an input attribute condition for dividing the analysis data group DA00 ′ into two so that the second data group DA2 is collected on the other side. Here, “the data group B is collected in the data group A” means that the probability of the correlation rule “if the data belongs to the data group B, the data belongs to the data group A” is high. To do.

  The input attribute conditions determined by the first input attribute condition determining unit 12 are as shown in Table 3. This input attribute condition is “the second data group DA2 of defective products in the data group that satisfies the input attribute condition”. ("Output attribute"> 2.5) is gathered, and the non-defective first data group DA1 ("Output attribute" ≤2.5) is gathered in a data group that does not satisfy the input attribute condition ". The input attribute condition (Table 3) determined by the first input attribute condition determination unit 12 is stored in the analysis result data storage unit 13.

  As a method for determining the input attribute condition by the first input attribute condition determining unit 12, any method for dividing the data group into two may be used. For example, the Gini index method described in Non-Patent Document 1 is used. Used.

  Hereinafter, a method of determining the input attribute condition by the Gini index method will be briefly described. In this Gini index method, an input attribute condition is determined using an index called “degree of improvement”.

  The “improvement degree Δi (t1)” of the Gini index method is “when the node t1 is branched into two child nodes t2 and t3 based on a certain input attribute condition, how much output attribute is caused by this branching. It is an index indicating whether or not the unity is improved, and is expressed by the following equation (1).

Δi (t1) = i (t1) − {pt2 · i (t2) + pt3 · i (t3)}
(1)
Here, pt2 and pt3 indicate branching ratios when the node t1 is branched into the child node t2 and the child node t3. Further, i (t1), i (t2), and i (t3) indicate Gini indexes at the node t1, the child node t2, and the child node t3, respectively.

  The Gini index i (tx) (x = 1, 2, 3) is an index representing the degree of grouping of output attribute values, and is represented by the following expression (2).

i (tx) = 1−Σ {p (j | tx)} ² (2)
Here, p (j | tx) indicates the probability that the output attribute is “j” at the node tx, and that the Gini index i (tx) is small indicates that the value of the output attribute is well organized at the node tx. It means that Therefore, an input attribute condition that increases the “improvement degree Δi (t1)” (that is, an input attribute condition that decreases i (t2) or i (t3)) sets the node t1 to improve the unity of output attributes. It can be said that it can be differentiated into two.

  The first input attribute condition determination unit 12 analyzes, for example, the first data group DA1 of non-defective products and the second data group DA2 of defective products together by using the improvement degree of the Gini index method as an index. An input attribute condition for dividing the data group DA00 ′ into two is determined.

  Specifically, with respect to all patterns of input attribute conditions that can be taken by each input attribute (for all branch patterns of data groups that satisfy the input attribute conditions and data groups that do not satisfy the input attribute conditions), the degree of improvement described above And the condition having the maximum improvement in each input attribute is extracted. Then, a condition corresponding to the second data group is determined as an input attribute condition among the two child nodes separated by the extracted condition.

  The determination of the input attribute condition in the first input attribute condition determination unit 12 is based on an index other than the improvement degree of the Gini index method, for example, “difference value between the first frequency and the second frequency” described later. May be based on. The determination method of the input attribute condition in the first input attribute condition determination unit 12 based on the above “difference value between the first frequency and the second frequency” is, for example, the first value for all the numerical values of each input attribute. In the data group DA1, the ratio of data whose input attribute is less than or equal to that value is calculated as the first frequency, and in the second data group DA2, the ratio of data whose input attribute is less than or equal to that value is calculated as the second frequency. Of all the numerical values taken by one input attribute, a frequency calculating step for calculating as a frequency, a difference calculating step for calculating a difference value between the first frequency and the second frequency for all numerical values of each input attribute The numerical value that maximizes the difference value is set as a threshold value for the input attribute, and a threshold value determining step for determining one threshold value for each input attribute and the threshold value determined in the threshold value determining step. Te may be a method comprising an input attribute condition determining step of determining the input attribute conditions.

[Step 5]
Among the input attribute conditions (Table 3) determined by the first input attribute condition determining unit 12, the branch condition determining unit 14 selects the input attribute condition having the highest probability of cutting out the second data group DA2 as the branch condition. Determine as. As an index of the accuracy with which the second data group DA2 is cut out, it is obtained by the improvement degree of the Gini index method, “the ratio of the second data group included in the data group corresponding to the input attribute condition”, or the frequency calculation step described above. A ratio between the first frequency and the second frequency, a difference value between the first frequency and the second frequency obtained by the frequency calculation step and the difference calculation step described above, a combination thereof, and the like may be used.

  In this example, the improvement degree of the Gini index method is used as an index of the accuracy of extracting the second data group DA2. In this way, “input attribute 6 ≦ 2.00” in Table 3 is determined as the branch condition.

[Step 6]
Based on the branch condition “input attribute 6 ≦ 2.00” determined by the branch condition determining unit 14, the data dividing unit 15 sets the analysis data group DA00 ′ to cause data that satisfies “input attribute 6 ≦ 2.00”. The group is divided into other data groups that do not satisfy “input attribute 6 ≦ 2.00” (that satisfies “input attribute 6> 2.00”). The other data group corresponds to “a data group having a smaller ratio of the second data group” in the claims.

[Step 7]
The analysis data group extraction unit 6 extracts only another data group from the data group divided by the data division unit 15 as the next analysis data group DA00 ′. Then, the series of processing from step 3 to step 6 is repeated until the end condition determination unit 16 determines that the end condition is satisfied. The end condition determination unit 16 of the present embodiment is configured to determine the end condition when the number of data in the second data group DA2 of defective products included in the analysis data group DA00 ′ is zero. As described above, detailed factor analysis results can be obtained by repeatedly performing the process until the number of data in the second data group DA2 of defective products becomes zero.

  The end condition is another end condition based on the number of data in the second data group DA2, for example, (1) when the number of data in the second data group DA2 is equal to or less than a predetermined number, (2) the first data group. When the ratio of the number of data in the second data group DA2 to the number of data in DA1 is equal to or less than a predetermined ratio, (3) the index of the extraction accuracy of the second data group DA2 under the branch condition determined in step 5 above is predetermined. It is good also when it falls below a value. When such termination conditions are used, a simpler and sufficient factor analysis result can be obtained. Furthermore, when priority is given to obtaining a simple factor analysis result, the end condition is simply set to the case where the iterative process is performed a predetermined number of times, or the end condition determining unit 16 is omitted to perform the iterative process as much as possible. Or you may.

  In the present embodiment (analysis of the basic data group DA in Table 1), the fourth analysis data group extracted by the analysis data group extraction unit 6 does not include the second data group DA2 of defective products. The process was completed after 3 repetitions. Table 4 shows the input attribute conditions determined for each input attribute by the first input attribute condition determination unit 12 (step 4) for each of the repetition processes.

[Step 8]
For each of the input attribute conditions (Table 4) determined by the first input attribute condition determination unit 12, the defective product data calculation unit 17 is in the classified basic data group DA00 (not the analysis data group DA00 ′). The number of data satisfying the input attribute condition ("DA1 + DA2" column in Table 5) and the number of data satisfying the input attribute condition and corresponding to the second data group DA2 of defective products ("Table 1" DA2 "column).

  The number of “DA2” columns represents the number of defects caused by each input attribute condition, and corresponds to the number of corresponding second data groups in the claims.

  Next, the defective product separation degree is calculated by dividing the value in the “DA2” column of Table 5 by the value in the “DA1 + DA2” column. The defective product separation degree of each input attribute condition is the ratio of the number of data in which the second data group DA2 which is a defective product is included in the data satisfying the input attribute condition (the input attribute condition in the classified basic data group DA00) The ratio of the number of data corresponding to the second data group DA2 of defective products among the data satisfying the above condition, the accuracy of the defective product extraction by the input attribute condition (the input attribute condition in the basic data group DA00 after classification) The defect rate when the data belonging to is a population), and corresponds to the second data group separation degree in the claims. This degree of defective product separation corresponds to the certainty of an association rule in the post-classification basic data group DA00 that “if the input attribute condition is satisfied, a defective product is generated”.

  Table 5 shows the total number of data (= 495: “DA1 + DA2” column) in the basic data group DA00 after classification in the “Total” row, along with the calculation result by the defective product data calculation unit 17, 12 is a table showing the number of second data groups DA2 (= 253: “DA2” column) and defective product content rate (= 253/495 = 0.511: “defective product separation” column). The defective product content rate is a ratio of the number of data in the second data group DA2 included in the basic data group DA00 after classification, and the defective rate when all the data in the basic data group DA00 after classification is a population ( The ratio of defective products). Hereinafter, the table of Table 5 is referred to as a first input attribute condition list table.

  The meaning of each column in Table 5 can be easily understood by referring to FIG. 5 in which these are represented by Venn diagrams for typical input attribute conditions (input attribute conditions extracted for the first time for input attribute 6 and input attribute 7). . FIG. 5A is a diagram showing the degree of defective product separation and the number of defects when defective products are cut out for the input attribute 6 based on the input attribute condition extracted for the first time, and FIG. It is a figure which shows the inferior goods isolation | separation degree and the number of inferiorities when performing the inferior goods cutting-out by the input attribute conditions extracted about the attribute 7 first time.

[Step 9]
The second input attribute condition determination unit 19 causes the problem event (second data group) among the plurality of input attribute conditions for the same input attribute in the first input attribute condition list table (Table 5). Select only input attribute conditions that have a high degree of importance as the factor of output attribute conditions corresponding to DA2, that is, a high degree of priority (hereinafter referred to as "priority") that should be taken when taking measures to eliminate the problem phenomenon To do.

  In the present embodiment, the second input attribute condition determination unit 19 performs a plurality of input attribute conditions for the same input attribute in the input attribute condition list table (Table 5), that is, n times in Step 7 (n is 2 or more). Among the n input attribute conditions determined by the (integer) repetitive processing, an input attribute condition that maximizes the defective product separation degree (second data group separation degree) is selected as an input attribute condition having a high priority. However, when the input attribute conditions of a plurality of types of patterns are included in the n input attribute conditions, the second input attribute condition determining unit 19 determines the defective product separation degree (second data group separation degree) for each pattern. The input attribute condition that maximizes is selected as an input attribute condition with high priority.

  Specifically, for the same input attribute, the input attribute condition that maximizes the defective product separation degree (second data group separation degree) in the first pattern “input attribute is below threshold” And one input attribute condition that maximizes the defective product separation degree (second data group separation degree) in the second pattern of “input attribute exceeds threshold”, and It is determined as information indicating the factor of the non-defective second data group DA2 (“output attribute”> 2.5).

  In the case of the example of Table 5, the 10 input attribute conditions shown in Table 6 were finally selected by the second input attribute condition determination unit 19.

  Here, the reason why the defective product separation degree (second data group separation degree) is used as an index in selecting the input attribute condition having a high priority is as follows. That is, as described above, the degree of defective product separation under a certain input attribute condition represents the accuracy of defective product extraction according to the input attribute condition. Therefore, an input attribute condition with a large value is a value other than other input attributes (disturbances). This is because it is considered to be a condition determined with high accuracy without being affected by the above.

  The second input attribute condition determination unit 19 may further perform the following processing. That is, in each of the input attribute conditions in the table of Table 6, a value larger than the defective product content rate of the basic data group DA00 after classification (= 0.511: “Defect Product Separation” column in the “Total” row). Select an input attribute condition that has a defect separation degree. In the example of Table 6, nine conditions shown in Table 7 were finally determined.

  The input attribute condition selected in this way is a condition including the second data group DA2 of defective products at a higher rate than the randomly selected sample from the classified basic data group DA00, and the second data group of defective products. The factor of DA2 is shown.

  As described above, the input attribute condition (Table 6 or Table 7) determined by the second input attribute condition determination unit 19 becomes the final input attribute condition output by the input attribute condition determination unit 100. Hereinafter, the table in Table 6 or 7 is referred to as a second input attribute condition list table.

  The process of the input attribute condition determination unit 100 (that is, the process of steps 3 to 9) is expressed in the form of a regression tree as shown in FIG. Referring to FIG. 6, in the present embodiment, among the two data groups (nodes) divided (branched) by the regression tree, “the data group having the smaller ratio of the defective second data group DA2”. Only the other data group side that is "is growing the regression tree sequentially. As described above, since the analysis (determination of input attribute conditions) is sequentially performed on the data group (other data group) in which the main cause of the defect (second data group DA2) is not found, the previous iterations are repeated. In the process, the influence of the input attribute condition that becomes the main cause of the defect is excluded, and a new factor can be extracted with high accuracy and efficiency. Moreover, the regression tree is not complicated and it is easy to intuitively grasp the cause of the failure.

  In the present embodiment, for each branch of the regression tree, not only the branch condition but also the input attribute condition that causes the problem event (defective product second data group DA2) is obtained for all the input attributes (first data). Of the input attribute condition determination unit 11). Then, among all the input attribute conditions for the number of times of branching (for the number of iterations), input attribute conditions with a higher degree of defective product separation are narrowed down and determined as the final failure factor (second input attribute) Processing by the condition determining unit 19). As described above, since the input attribute condition that causes the problem event (defective product second data group DA2) is extracted for all input attributes, not only the branch condition in the regression tree, the competing factor is included in the branch condition. Even if there is, can be caught without missing the factor.

  In the final factor determination (the final input attribute condition determination process by the second input attribute condition determination unit 19), the factor is determined based on a clear index of defective product separation. Even if it is a complicated regression tree, the cause of the failure can be clearly grasped.

  As described above, the second input attribute condition determination unit 19 uses a plurality of the same input attributes determined for each repetition process based on a clear index of defective product separation (second data group separation). Since only high priority conditions are selected from among the input attribute conditions, the above process is repeated in order to obtain a detailed cause analysis result. The cause of the output attribute condition corresponding to the second data group can be determined with high accuracy.

[Step 10]
The input attribute condition list graph creation unit 20 creates an input attribute condition list graph (FIG. 7) based on the second input attribute condition list table (Table 7). In the input attribute condition list graph (FIG. 7), the number of defects (the number corresponding to the second data group) for each of the input attribute conditions determined by the second input attribute condition determining unit 19 (input attribute condition determining unit 100) is displayed. The bar graph and the defective product separation degree (second data group separation degree) are each shown by a line graph. The left end (“Total”) of the graph shows the number of defects (= 253) and the defective product content rate (= 0.511) in the basic data group DA00 after classification. In FIG. 7 and other drawings, the numerical value in the form “... E...” Is a numerical value in decimal exponent notation, the number before E is the mantissa, and the number after E. Is the exponent part.

  The horizontal axis of the input attribute condition list graph (FIG. 7) shows each input attribute condition (each input attribute condition in Table 7) determined by the second input attribute condition determination unit 19. The order of the input attribute conditions is as follows: (1) in order of increasing defective product separation (second data group separation), (2) number of defects (number of second data groups DA2: number of second data groups) Rearrangement is possible in the descending order, and (3) in the descending order of the score considering both the degree of defective product separation and the number of defects. Here, the “scoring in consideration of both the defective product separation degree and the number of defects” means the first scoring given so that the score increases in descending order of the defective product separation degree (second data group separation degree). The total number of points with the second score given so that the score increases in descending order of the number of defects (corresponding number of second data group).

  8 to 10 are respectively (1) in order of increasing defective product separation, (2) in order of increasing number of defects, and (3) in order of increasing score considering both the degree of defective product separation and the number of defects. The horizontal axis (input attribute conditions) is rearranged, and the input attribute condition list graph is shown. Hereinafter, the meaning of the order of the above (1) to (3) will be described.

(1) When the purpose of analysis is to separate defective products with high accuracy by separating them from good products having a large defective product separation degree (second data group separation degree), importance is attached to the defective product separation degree. The user can resolve the defect with high accuracy without dropping the non-defective product as much as possible by taking countermeasures in order from the input attribute condition having a large degree of defective product separation. However, in the input attribute condition with a large degree of defective product separation, the number of defects is not necessarily large (if it is a decisive failure factor but its frequency of occurrence is low), and even if measures are taken, There is a risk that only a part of the whole defect is resolved.

(2) In order of increasing number of defects (corresponding to the number of second data group) When the purpose of analysis is to reduce the number of defects currently occurring as much as possible, the number of defects is emphasized. The user can resolve many of the currently occurring defects by taking countermeasures in order from the input attribute condition with the largest number of defects. However, in the case of input attribute conditions with a large number of defects, the degree of defective product separation is not necessarily large (when the frequency of occurrence is high, but not necessarily a decisive factor of failure). At the same time, there is a risk of dropping good products.

(3) High order of distribution considering both the degree of defective product separation and the number of defectives The purpose is to reduce many defects with high accuracy, considering both the degree of defective product separation and the number of defective products. In this case, the emphasis is on “scoring considering both the degree of defective product separation and the number of defects”. The user can resolve many defects with high accuracy by taking measures in descending order of the score. However, the effect of “resolving defects without dropping good products” by this measure is inferior to (1), and the effect of “removing many defects” is inferior to (2).

  The user operates the “item rearrangement” button in FIG. 7 according to the purpose of analysis, selects one of the above-mentioned (1) to (3), and changes the order of input attribute conditions. change. Then, a review graph (in step 11), which will be described later, is created in this order of arrangement, and technical analysis is added to review the analysis results. In other words, the input attribute condition list graphs (FIGS. 7 to 10) serve as a navigation that informs the user of the provisional order of the analysis result review.

  The input attribute condition list graph creation unit 20 uses the order determination unit 23 to arrange the input attribute conditions in accordance with the order selected by the user (any one of (1) to (3)) as follows. Done.

  First, according to the purpose of analysis, the user instructs the data analysis apparatus which one of the three types of arrangement orders (1) to (3) is to be adopted. In response to this, the order determining unit 23 prioritizes the input attribute conditions corresponding to the instructed order. That is, for each input attribute condition (each input attribute condition in the second input attribute condition list table (Table 7)) determined by the rank determining unit 23 as the cause of the problem phenomenon by the second input attribute condition determining unit 19 In order to resolve the problem event, the ranking is based on the importance (priority) as the cause of the problem event (factor causing the output attribute condition corresponding to the second data group DA2 of the defective product). The order (priority order) that should be prioritized is assigned (prioritization is performed). Next, the input attribute condition list graph creating unit 20 creates an input attribute condition list graph in which the input attribute conditions are arranged according to the ranking result by the rank determining unit 23.

  Here, the arrangement order of the input attribute conditions in the input attribute condition list graph can be selected from the three arrangement orders (1) to (3), but two of these three arrangements are arranged. It may be selectable from the order or may not be selectable. However, it is preferable that the selection can be made from the three types of arrangement orders (1) to (3) in that the user can easily select the input attribute condition according to the purpose of the analysis.

  Here, the input attribute condition list graph creation unit 20 shows the ranking result of the rank determination unit 23 to the user as the order of the input attribute conditions in the input attribute condition list graph. The condition list graph creation unit 20 may show the result of ranking by the rank determination unit 23 to the user in another form. For example, the input attribute condition list graph creating unit 20 simply creates an input attribute condition list graph indicating a list of input attribute conditions, and sets the priority of each input attribute condition determined by the rank determining unit 23 to the input attribute condition list. It may be shown in a graph.

[Step 11]
Referring to the order (priority order) of multiple input attribute conditions shown in the input attribute condition list graph, the user can select a specific input attribute from the multiple input attribute conditions shown in the input attribute condition list graph. A selection operation for selecting a condition and a review graph drawing instruction are performed.
The selection operation by the user is an operation by the input device 31 on the display screen of the input attribute condition list graph (FIGS. 7 to 10), for example, a character indicating a specific input attribute condition on the display screen of the input attribute condition list graph Is performed by an operation of pointing (pointing) with the pointing device (mouse or the like) as the input device 31. Further, in this case, the review graph drawing instruction by the user is an operation by the input device 31 on the display screen of the input attribute condition list graph, for example, for creating the review graph displayed on the display screen of the input attribute condition list graph. This is performed by an operation of pointing (pointing) a button with a pointing device (such as a mouse) as the input device 31. Note that the selection operation and review graph drawing instruction by the user can be performed by an operation other than the operation on the display screen of the input attribute condition list graph, for example, an operation of inputting a specific character using the keyboard as the input device 31. However, it is preferable that the operation can be performed by the input device 31 on the display screen of the input attribute condition list graph because the operation by the user is easy. Note that an operation for selecting two or more specific input attribute conditions described later can also be performed by the above-described operation.

  Then, in response to the selection operation and the review graph drawing instruction, the review graph creation unit 21 acquires the selected input attribute condition as a specific input attribute condition, and data that satisfies the acquired specific input attribute condition A review graph that visually represents the correspondence between the group and the second data group is created.

  When each of the data constituting the basic data group belongs to any of a plurality of groups, the review graph creating unit 21 responds to the selection operation and the review graph drawing instruction in response to the selection operation and the review graph drawing instruction. In each unit, a second basic data group representing a data group obtained by averaging the values of the input attribute and output attribute belonging to the group is created, and the selected input attribute condition is used as a specific input attribute condition. A data group that satisfies the specific input attribute condition acquired in the second basic data group, and a data group that satisfies the output attribute condition corresponding to the second data group in the second basic data group; It is also possible to create a review graph that visually represents the correspondence relationship.

  The specific input attribute condition acquired by the review graph creating unit 21 is not limited to one and may be plural. When there are a plurality of input attribute conditions to be acquired, the review graph creating unit 21 creates review graphs corresponding to the number of acquired input attribute conditions, and a graph of these lists or a graph stored in a page format. A review graph is output as a group.

  In addition, the method in which the review graph creation unit 21 acquires a specific input attribute condition for creating the review graph is not limited to the selection operation by the user but may be automatic selection by a computer. For example, the computer automatically selects a predetermined number of input attribute conditions in the descending order of the degree of defective product separation, or automatically selects all the input attribute conditions shown in the input attribute condition list graph. Also good.

  Further, the review graph drawing instruction to the review graph creating unit 21 is not limited to the drawing instruction by the user, but may be an instruction from a computer based on preset contents.

The review graph created by the review graph creation unit 21 is, for example, as shown in FIG. 11, (1) trend graph (first review graph), (2) map (second review graph), (3) scatter There are four types: a diagram (third review graph) and (4) a frequency distribution diagram (fourth review graph).
Hereinafter, in conjunction with the analysis result review method based on the input attribute condition list graph (FIG. 8) arranged in descending order of defective product separation (second data group separation), the above (1) to (4) The review graph will be described.

  Note that the review graphs (1) to (4) include, for example, a button for creating a review graph on the display screen of the input attribute condition list graph (FIG. 8) (“trend / map / scatter diagram” button and In response to operating any one of the “frequency distribution map” buttons) with the pointing device as the input device 31, the review graph creating unit 21 creates it.

(1) Trend graph (first review graph)
The review graph creation unit 21, for example, in response to the user selecting a specific input attribute condition from the input attribute conditions shown in the input attribute condition list graph and instructing the trend graph drawing, A first review graph is created as a review graph. The first review graph includes a first trend graph showing a time change of an input attribute corresponding to a specific input attribute condition selected by a user, and a second trend graph showing a time change of the output attribute. Both are shown. In the first trend graph, marking processing is performed on data satisfying the input attribute condition and satisfying the output attribute condition corresponding to the second data group.

  For example, the review graph creation unit 21 is configured so that the user can select an input attribute condition (input attribute 6 ≦ 2.00) having the highest priority among the input attribute conditions shown in the input attribute condition list graph (FIG. 8). , Input attribute 2 ≦ 2.01, input attribute 7> 1.15, input attribute 8> 10.7, input attribute 3> 2.00, input attribute 4> 2.03), and a trend graph drawing instruction In response to this, the trend graph shown in FIG. 12 is created, the trend graph is output to the display device 32, and the trend graph is drawn on the screen of the display device 32. For example, after the user operates the “trend / map / scatter diagram” button on the display screen of the input attribute condition list graph (FIG. 8) with the pointing device as the input device 31, Is selected by the input device 31.

  In the trend graph of FIG. 12, the time change of each input attribute (input attribute 6, input attribute 2, input attribute 7, input attribute 8, input attribute 3, input attribute 4) corresponding to the selected six input attribute conditions And a second trend graph showing the time change of the output attribute are displayed side by side. In the trend graph of FIG. 12, the second trend graph is displayed side by side with the first trend graph. However, the second trend graph is displayed in a color different from the first trend graph and the first trend graph. It may be displayed in an overlapping manner. In these graphs, the average value of the “input attribute” or “output attribute” value of each element in the same substrate is shown for each substrate. The horizontal axis indicates the substrate number. In this example, this arrangement order corresponds to the manufacturing order of the elements (substrates).

  Here, in each trend graph (first trend graph) of each input attribute, the data satisfying the input attribute condition of the input attribute and satisfying the output attribute condition corresponding to the second data group DA2 of the defective product. Is subjected to a red marking process (white on the drawing). In the output attribute trend graph (second trend graph), data that satisfies the output attribute condition corresponding to the second data group DA2 of the defective product is subjected to red marking (outlined in the drawing). . That is, in the trend graph of FIG. 12, the correspondence between the defective second data group DA2 and the data group that satisfies the input attribute condition extracted as the factor is visually highlighted.

  Therefore, by referring to the trend graph of FIG. 12, the user can change the failure factor (input attribute condition) determined by the second input attribute determining unit 19 (input attribute condition determining unit 100) for each board. It is possible to immediately understand the synchrony with the output attribute in (time change). Specifically, in the periods A and B in FIG. 12, the condition that the input attribute 6 and the input attribute 2 are low (input attribute condition “input attribute 6 ≦ 2.00” and “input attribute 2 ≦ 2.01”). ”(Input attribute condition“> ”) is caused by a condition (input attribute condition“ input attribute 7> 1.15 ”). It can be seen that defects are caused mainly by the condition that the input attribute 4 is high in the period D (input attribute condition “input attribute 4> 2.03”). In addition, the change in the microscopic output attribute throughout the entire period (the circled portion) may be the influence of the condition where the input attribute 3 is high (the input attribute condition “input attribute 3> 2.00”). I understand. On the other hand, the input attribute 8 does not have a clear correlation with the output attribute, but this point will be clarified in the description of the map (FIG. 13) described later.

  It should be noted that the horizontal axis of the trend graph can be arranged in order from the smallest number using a substrate number or a lot number. Moreover, it is also possible to arrange the trend graph in order from the oldest processing date and time using the processing date and time in a specific process for manufacturing or inspecting elements on the horizontal axis of the trend graph. That is, it can be arranged with reference to a discrimination key (substrate number, lot number, processing date and time) corresponding to each data. In the present invention, a graph with the horizontal axis representing the arrangement based on the discrimination keys as described above is called a trend graph.

  In the above, the basic data group in Table 1 is an analysis target, and each piece of data is an element number unit (that is, each element belongs to both a specific board number and a specific element coordinate number). Therefore, the trend graph was created using the average value of the “input attribute” or “output attribute” value of each element on the same board. There is no need to use it.

(2) Map (second review graph)
For example, the review graph creation unit 21 responds to the fact that the user selects a specific input attribute condition from the input attribute conditions shown in the input attribute condition list graph and gives a map drawing instruction, in response to the review drawing. A second review graph is created as a graph. The second review graph includes both a first map representing an in-plane distribution of input attributes corresponding to the selected specific input attribute condition, and a second map representing an in-plane distribution of the output attributes. Show. In the first map, data that satisfies the input attribute condition and satisfies the output attribute condition corresponding to the second data group is subjected to marking processing.

  The above map (second review graph) is a graph representing changes in input attributes and output attributes for each discrimination key, using the position (element coordinate number) in the substrate surface as the discrimination key.

  For example, the review graph creation unit 21 is a user representative of the six input attribute conditions having the highest priority among the input attribute conditions shown in the input attribute condition list graph (FIG. 8). In response to selecting two input attribute conditions (input attribute 6 ≦ 2.00, input attribute 8> 10.7) and instructing map drawing, the map shown in FIG. 13 is created and the map is displayed. Output to the device 32 and draw on the screen of the display device 32. For example, after the user operates the “trend / map / scatter chart” button on the display screen of the input attribute condition list graph (FIG. 8) with the pointing device as the input device 31, the map drawing instruction is further displayed as “map”. Is selected by the input device 31.

  In the map of FIG. 13, a first map representing an in-plane distribution (change due to position on the substrate) of each input attribute (input attribute 6, input attribute 8) corresponding to two selected input attribute conditions; A second map representing the in-plane distribution of output attributes is displayed side by side. In these maps, the value of “input attribute” or “output attribute” of each element in the same “element coordinate number” is displayed. Are averaged for each element coordinate position. Specifically, in the output attribute map (second map), the coordinate position corresponding to the condition (“output attribute”> 2.5) where the average value of the output attribute is defective is red (shaded in the drawing). The other coordinate positions are shown in green (indicated by hatching in the drawing). In this example, the average value of output attributes at coordinate positions P1 and P2 is 2.71, the average value of output attributes at coordinate position P3 is 2.46, and the average value of output attributes at coordinate positions P4 and P5 is 2.22. Therefore, the coordinate positions P1 and P2 are shown in red (indicated by hatching in the drawing). In each input attribute map (first map), the average value of the input attributes of each element corresponding to the coordinate position satisfies the input attribute condition (input attribute 6 ≦ 2.00, input attribute 8> 10.7). The coordinate position is shown in red (indicated by hatching in the drawing), while the coordinate position where the average value of the input attribute of each element corresponding to the coordinate position does not satisfy the input attribute condition is green ( This is indicated by hatching in the drawing). In this example, since the average values of the input attributes 6 at the coordinate positions P1 to P5 are all 1.96 (≦ 2.00), in the first map related to the input attributes 6, the coordinate positions P1 to P5 are red (drawing). It is shown in the shaded hatch above). In this example, the average value of the input attribute 8 at the coordinate positions P1 and P2 is 11.5, the average value of the input attribute 8 at the coordinate position P3 is 10.5, and the average value of the input attribute 8 at the coordinate position P4 is Since the average value of the input attribute 8 at 9.52 and the coordinate position P5 is 9.51, the coordinate positions P1 and P2 are red (indicated by hatching in the drawing) in the first map related to the input attribute 6. It is shown in

Here, in each of the two first maps representing the in-plane distribution of each input attribute, the input attribute condition of the input attribute is satisfied, and the output attribute condition corresponding to the second data group DA2 of the defective product is satisfied. The data is marked with a red frame (black frame in the drawing). That is, in the map of FIG. 13, the correspondence between the second data group DA2 of defective products and the data group that satisfies the input attribute condition extracted as the factor is visually emphasized.
Therefore, by referring to the map of FIG. 13, the user can determine the failure factor (input attribute condition) determined by the second input attribute determination unit 19 (input attribute condition determination unit 100) in the in-plane tendency. Immediate understanding of synchrony with output attributes. Specifically, it can be seen that the defect of the element coordinate numbers 1 and 2 (“output attribute”> 2.5) in FIG. 13 is mainly caused by the higher input attribute 8. On the other hand, the input attribute 6 has no clear correlation with the output attribute. Considering this result together with the trend graph (FIG. 12), the input attribute 6 is a factor that affects the output attribute in units of each board, whereas the input attribute 8 is a surface in the board. It can be seen that this is a factor that affects the output attribute as an internal distribution.

  In the above, the basic data group in Table 1 is an analysis target, and each piece of data is an element number unit (that is, each element belongs to both a specific board number and a specific element coordinate number). Therefore, a map was created using the average value of the “input attribute” or “output attribute” value of each element in the same “element coordinate number”. For example, it is not necessary to use the average value.

(3) Scatter chart (third review graph)
In response to the user selecting a specific input attribute condition from the input attribute conditions shown in the input attribute condition list graph and giving a scatter diagram drawing instruction, for example, the review graph creation unit 21 A third review graph is created as a review graph. The third review graph is a scatter diagram showing the relationship between the input attribute corresponding to the selected specific input attribute condition and the output attribute. In this scatter diagram, marking processing is applied to data satisfying the input attribute condition and satisfying the output attribute condition corresponding to the second data group.

  For example, the review graph creation unit 21 selects the input attribute condition (input attribute 6 ≦ 2.00) having the highest priority among the input attribute conditions shown in the input attribute condition list graph (FIG. 8). In response to the scatter diagram drawing instruction, a scatter diagram showing the relationship between the input attribute (input attribute 6) corresponding to the selected input attribute condition and the output attribute is created and the scatter diagram is displayed. Output to the device 32 and draw a scatter diagram on the screen of the display device 32. For example, after the user operates the “trend / map / scatter diagram” button on the display screen of the input attribute condition list graph (FIG. 8) with the pointing device as the input device 31, This is done by selecting “Figure” with the input device 31.

  In the scatter diagram, each data is indicated by “◯”. Here, in the above scatter diagram, the data that satisfies the input attribute condition corresponding to the input attribute on the horizontal axis and that satisfies the output attribute condition corresponding to the second data group DA2 of the defective product is red (on the drawing). (In gray) marking processing is performed (that is, data satisfying the input attribute condition and satisfying the output attribute condition corresponding to the defective second data group DA2 is represented by a red dot (gray in the drawing). Indicated by a dot). That is, in the scatter diagram of FIG. 14, the correspondence between the defective second data group DA2 and the data group that satisfies the input attribute condition extracted as the factor is visually emphasized.

  Therefore, by referring to the scatter diagram of FIG. 14, the user can evaluate the failure factor (input attribute condition) determined by the second input attribute condition determination unit 19 (input attribute condition determination unit 100) in a multivalued manner. You can immediately understand the correlation with the output attributes. Specifically, it is intuitive that the low condition of the input attribute 6 (input attribute 6 ≦ 2.00) affects the defect (“output attribute”> 2.5) even when viewed in multiple values. Can understand.

  The review graph creating unit 21 responds to the fact that the user has instructed the grouping by another input attribute condition different from the input attribute condition corresponding to the axis of the scatter diagram, and satisfies the other input attribute condition. It is also possible to show data and data that does not satisfy the other input attribute conditions in different forms in the scatter diagram. In this case, the user performs a selection operation for selecting a specific input attribute condition as an axis of the scatter diagram and a scatter diagram drawing instruction at the stage of the scatter diagram drawing instruction, and is different from the input attribute condition. It suffices to indicate the grouping according to the input attribute condition. FIG. 15 shows the result of instructing the grouping by the input attribute 7> 1.15 at the stage of drawing instruction of the scatter diagram of FIG. In FIG. 15, data satisfying another input attribute condition “input attribute 7> 1.15” is indicated by “x”, while data not satisfying the other input attribute condition “input attribute 7> 1.15” is represented by “ “○”. By referring to this result, it can be seen that the input attribute 7> 1.15 (x in the figure) is a factor of failure independently of the input attribute 6, and data excluding this influence In (input attribute 7 ≦ 1.15), the relationship between the input attribute 6 and the output attribute can be confirmed more remarkably.

(4) Frequency distribution chart (fourth review graph)
In response to the user selecting a specific input attribute condition from the input attribute conditions shown in the input attribute condition list graph and instructing a frequency distribution diagram, for example, The basic data group is separated into a data group that satisfies the selected specific input attribute condition and a data group that does not satisfy the input attribute condition, and shows the frequency distribution of the output attribute in each of the data groups A fourth review graph is created as the review graph.

  For example, the review graph creation unit 21 selects the input attribute condition (input attribute 6 ≦ 2.00) having the highest priority among the input attribute conditions shown in the input attribute condition list graph (FIG. 8). In response to the instruction to draw the frequency distribution diagram, the frequency distribution diagram shown in FIG. 16 is created, the frequency distribution diagram is output to the display device 32, and the frequency distribution diagram is drawn on the screen of the display device 32. To do. The frequency distribution diagram drawing instruction is performed, for example, when the user selects the “frequency distribution” button on the display screen of the input attribute condition list graph (FIG. 8) with the pointing device as the input device 31. The frequency distribution diagram is a line graph in FIG. 16, but may be a bar graph in which the frequency for each section on the horizontal axis is represented by a section width bar, that is, a so-called histogram.

  In the frequency distribution diagram of FIG. 16, the basic data group DA00 is divided into a data group that satisfies the selected input attribute condition (input attribute 6 ≦ 2.00) and an input attribute condition that does not satisfy the input attribute condition (input attribute 6> 2.00). The frequency distribution of output attributes in each data group when separated into data groups is shown. In the figure, output attribute threshold values (2.5) are shown when the basic data group is classified into a good first data group DA1 and a defective second data group DA2. That is, the frequency distribution diagram of FIG. 16 shows the correspondence between the defective second data group DA2 and the data group satisfying the input attribute condition extracted as the factor, and the non-defective first data group DA1 and the input attribute. A correspondence relationship with a data group that does not satisfy the condition is shown.

  By referring to the frequency distribution diagram of FIG. 16, the user can determine the failure factor (input attribute condition) determined by the second input attribute determination unit 19 (input attribute condition determination unit 100) and the exclusive condition. You can intuitively understand the degree of separation between good and defective products.

[Step 12]
The review graphs (FIGS. 12 to 16) mainly review the relationship between the individual input attribute conditions of the input attribute condition list graphs (FIGS. 7 to 10) and the output attributes. However, when a plurality of input attribute conditions are entangled with each other or in a master-slave relationship, it is necessary to evaluate an analysis result based on a condition in which a plurality of input attribute conditions are entangled. Therefore, a composite factor graph is created by the composite factor graph creation unit 22, and an analysis result is evaluated based on a condition in which a plurality of input attribute conditions are entangled using the composite factor graph.

  The composite factor graph creation unit 22 responds to a selection operation for selecting a specific input attribute condition and a graph drawing instruction on the screen of the input attribute condition list graph (FIGS. 7 to 10). A plurality of input attribute conditions including the selected input attribute condition are acquired as two or more specific input attribute conditions, and a composite based on a condition in which the acquired two or more specific input attribute conditions are entangled Create a factor graph.

  Note that the method by which the composite factor graph creation unit 22 acquires specific input attribute conditions for creating the composite factor graph is not limited to the selection operation by the user, but may be automatic selection by a computer. For example, the computer automatically selects a predetermined number of input attribute conditions in the descending order of the degree of defective product separation, or a combination of a predetermined number of input attribute conditions from among the input attribute conditions shown in the input attribute condition list graph All of these may be automatically selected.

  Further, the graph drawing instruction to the composite factor graph creating unit 22 is not limited to the drawing instruction by the user, but may be an instruction from a computer based on preset contents.

  The complex factor graph created by the complex factor graph creating unit 22 includes, for example, as shown in FIG. 11, (5) a complex factor Pareto diagram (first complex factor graph), (6) factor accumulated graph (second complex graph). Factor graph) and (7) best case / worst case frequency distribution diagram (third composite factor graph). Hereinafter, together with the evaluation method of the confounding condition based on the input attribute condition list graph (FIG. 8) arranged in descending order of the defective product separation degree (second data group separation degree), the above (5) to (7) The composite factor graph will be described. Note that the composite factor graphs (5) to (7) are, for example, that the user operates a “complex factor analysis” button or a “factor cumulative graph” button on the screen of the input attribute condition list graph (FIG. 8). In response to this, it is created by the composite factor graph creation unit 22.

(5) Composite factor Pareto chart (first composite factor graph)
For example, the composite factor graph creation unit 22 responds to the user selecting a specific input attribute condition from the input attribute conditions shown in the input attribute condition list graph and instructing the drawing of the composite factor Pareto diagram. Then, two specific input attribute conditions are acquired, and a first composite factor graph is created as the composite factor graph. The first composite factor graph satisfies the above two input attribute conditions in the basic data group and calculates the number of data corresponding to the second data group, and in the basic data group. FIG. 11 is a complex factor Pareto diagram showing the results of calculating the number of data satisfying only one of the two input attribute conditions and corresponding to the second data group.

  For example, the composite factor graph creation unit 22 has an input attribute condition (input attribute 6 ≦ 2.00) having the highest priority among the input attribute conditions shown in the input attribute condition list graph (FIG. 8). Is selected as a key condition, and a composite factor graph shown in FIG. 17 is created in response to the drawing instruction of the composite factor Pareto diagram, and the composite factor graph is output to the display device 32. Draw a composite factor graph on the screen. The drawing instruction of the complex factor Pareto diagram is, for example, after the user operates the “complex factor analysis” button on the display screen of the input attribute condition list graph (FIG. 8) with the pointing device as the input device 31, This is done by selecting the “factor Pareto diagram” with the input device 31.

  The complex factor Pareto diagram in FIG. 17 shows that each input attribute condition in FIG. 8 is defective (indicated by hatching in the drawing) that overlaps with the key condition (input attribute 6 ≦ 2.00). The number of defects (bar graph) is shown separately for the bar) and key conditions (input attribute 6 ≦ 2.00) and non-overlapping defects (bars in gray in the figure (shown without diagonal lines in the drawing)). Is. The number of defective key conditions is indicated by a red bar (indicated by shading in the drawing). Here, the number of defects overlapping with the key condition represents the number of data satisfying both of the two input attribute conditions including the key condition and corresponding to the second data group, and the number of defects not overlapping with the key condition is 2 This represents the number of data satisfying only one of the two input attribute conditions (conditions that are not key conditions) and corresponding to the second data group. In the composite factor Pareto diagram of FIG. 17, the degree of defective product separation is shown as a line graph. Note that the line graph of the degree of defective product separation may be omitted.

  Note that when creating the complex factor Pareto diagram of FIG. 17, one of the two specific input attribute conditions acquired by the complex factor graph creating unit 22 is the key condition (input attribute 6 ≦ 2.00). ), And another one for each input attribute condition other than the key condition.

  By referring to the complex factor Pareto diagram of FIG. 17, the user can select the key condition (input attribute) among the failure factors (input attribute conditions) determined by the second input attribute determining unit 19 (input attribute condition determining unit 100). It is possible to search for a factor independent of the complex factor (or the factor in the master-slave relationship) with the attribute 6 ≦ 2.00 and the key condition (input attribute 6 ≦ 2.00). Specifically, the defects caused by the input attribute 2 ≦ 2.01 in FIG. 17 include most of the defects of the key condition (input attribute 6 ≦ 2.00), and the key condition (input attribute 6 ≦ 2. It can be seen that this is a factor having a master-slave relationship with (00). This result corresponds to the input attribute 2 and the input attribute 6 changing in conjunction with each other in the trend graph of FIG.

  On the other hand, the defect caused by the input attribute 7> 1.15 in FIG. 17 has a relatively large proportion of defects that do not overlap with the key condition (input attribute 6 ≦ 2.00), and the key condition (input attribute 6 ≦ 2.00). It can be seen that this is an independent factor. As a result, in the trend graph of FIG. 12, the failure caused by the low input attribute 6 (input attribute 2) (input attribute 6 ≦ 2.00 (input attribute 2 ≦ 2.01)) This corresponds to the fact that a defect caused by a high input attribute 7 (input attribute 7> 1.15) occurs in a period C different from the above periods A and B. To do.

  Since the composite factor Pareto diagram (FIG. 17) corresponds to the result of the trend graph as described above, first, a composite factor Pareto diagram is created and the input attribute condition to be reviewed with reference to this is shown. The candidates may be listed and then the trend graph of FIG. 12 may be created.

(6) Factor accumulation graph (second composite factor graph)
In order to confirm the validity of these input attribute conditions for a plurality of input attribute conditions selected by the user with reference to the arrangement order of the input attribute condition list graph (FIG. 8) and the composite factor Pareto diagram (FIG. 17), That is, in order to confirm whether the failure factor can be selected without omission, the complex factor graph creating unit 22 creates a factor accumulation graph (second complex factor graph).

  In the composite factor graph creation unit 22, for example, the user selects a plurality of input attribute conditions from among the input attribute conditions indicated in the input attribute condition list graph, and gives a factor accumulation graph drawing instruction. In response, the selected input attribute conditions are ranked and acquired, and a second composite factor graph is created as the composite factor graph. In the second composite factor graph, the number of data satisfying the input attribute condition in the basic data group and corresponding to the second data group in the order of the order of the input attribute condition acquired by assigning the rank. It is the factor accumulation graph which accumulated sequentially.

  Note that the method by which the composite factor graph creation unit 22 acquires specific input attribute conditions for creating the factor accumulation graph is not limited to the selection operation by the user, but may be automatic selection by a computer. For example, the computer may automatically select a predetermined number of input attribute conditions in the descending order of the degree of defective product separation in the order of the order.

  For example, the composite factor graph creating unit 22 selects a plurality of input attribute conditions in a ranking order, for example, among the input attribute conditions shown in the input attribute condition list graph (FIG. 8). Six input attribute conditions within the top six priorities (input attribute 6 ≦ 2.00, input attribute 2 ≦ 2.01, input attribute 7> 1.15, input attribute 8> 10.7, input attribute 3> 2.00, input attribute 4> 2.03) are selected in that order, and in response to issuing a graph drawing instruction, the factor accumulation graph shown in FIG. 18 is created, and the factor accumulation graph is displayed on the display device 32. And a factor accumulation graph is drawn on the screen of the display device 32. The factor accumulation graph drawing instruction is performed, for example, when the user operates the “factor accumulation graph” button on the display screen of the input attribute condition list graph (FIG. 8) with the pointing device as the input device 31. In the factor accumulation graph of FIG. 18, the degree of defective product separation is shown as a line graph. Note that the line graph of the degree of defective product separation may be omitted.

  In the factor accumulation graph of FIG. 18, a pink bar (indicated by a diagonal line in the drawing) in a certain input attribute condition indicates a defect due to the input attribute condition on the left side of the number of defects in the input attribute condition. These are the number of defects that do not overlap with each other, and are stacked stepwise for each input attribute condition to express the total number of defects due to a plurality of input attribute conditions. An orange bar (shown without a diagonal line in the drawing) represents the number of defects that overlap with any of the defects caused by the input attribute condition on the left side.

  By referring to the factor accumulation graph of FIG. 18, the user can confirm the validity of the selected plurality of input attribute conditions, that is, whether the failure factor can be selected without omission. Specifically, the top three input attribute conditions selected by the user (input attribute 6 ≦ 2.00, input attribute 2 ≦ 2.01, input attribute 7> 1.15) are about 80% of the total number of defects. And the top five input attribute conditions (input attribute 6 ≦ 2.00, input attribute 2 ≦ 2.01, input attribute 7> 1.15, input attribute 8> 10.7, input attribute 3 > 2.00) account for almost all of the total number of defects.

(7) Best case / worst case frequency distribution chart (third composite factor graph)
For example, the composite factor graph creation unit 22 selects a plurality of input attribute conditions from the input attribute conditions shown in the input attribute condition list graph, and instructs the drawing of the best case / worst case frequency distribution diagram. In response to this, the selected input attribute condition is acquired, and a third composite factor graph is created as the composite factor graph. The third composite factor graph includes a data group that satisfies all of the two or more acquired input attribute conditions and an exclusive condition for each of the two or more acquired input attribute conditions from the basic data group. FIG. 6 is a best case / worst case frequency distribution diagram showing a frequency distribution of output attributes in each of the data groups, in which all data groups that satisfy all are extracted.

  Note that the method by which the composite factor graph creation unit 22 acquires specific input attribute conditions to create the best case / worst case frequency distribution diagram is not limited to the selection operation by the user, but is automatic selection by a computer. May be. For example, the computer may automatically select a predetermined number of input attribute conditions in the descending order of the degree of defective product separation.

  For example, the composite factor graph creation unit 22 selects a plurality of input attribute conditions, for example, among the input attribute conditions shown in the input attribute condition list graph (FIG. 8), the priority is higher. In response to selecting two typical input attribute conditions (input attribute 6 ≦ 2.00, input attribute 7> 1.15) and instructing drawing of the best case / worst case frequency distribution diagram, 19 is created, the best case / worst case frequency distribution diagram is output to the display device 32, and the best case / worst case frequency distribution diagram is drawn on the screen of the display device 32. To do. The drawing instruction of the best case / worst case frequency distribution chart is, for example, after the user operates the “complex factor analysis” button on the display screen of the input attribute condition list graph (FIG. 8) with the pointing device as the input device 31. Further, the “best case / worst case frequency distribution diagram” is selected by the input device 31. The best case / worst case frequency distribution diagram is a line graph in FIG. 19, but may be a bar graph in which the frequency for each section on the horizontal axis is represented by a section width bar, that is, a so-called histogram.

  In the best case / worst case frequency distribution diagram of FIG. 19, a data group satisfying all the selected input attribute conditions (input attribute 6 ≦ 2.00 and input attribute 7> 1.15: worst case data group. And a data group (input attribute 6> 2.00 and input attribute 7 ≦ 1.15: best case) satisfying all exclusive conditions for each of the selected input attribute conditions. A frequency distribution diagram of output attributes in each data group when extracted from the data group is drawn. In FIG. 19, the best case frequency is described as “best case frequency”, and the worst case frequency is described as “worst case frequency”. In the figure, output attribute threshold values (2.5) are shown when the basic data group is classified into a good first data group DA1 and a defective second data group DA2. That is, the frequency distribution diagram of FIG. 19 shows the mutual influence of a plurality of input attribute conditions extracted as factors of the defective second data group DA2.

  By referring to the best case / worst case frequency distribution diagram of FIG. 19, the user can confirm the influence between a plurality of input attribute conditions of interest. Specifically, the condition of the input attribute 6 ≦ 2.00 and the input attribute 7> 1.15 is greater than the case where the input attribute 6 ≦ 2.00 is acting alone (FIG. 16). You can see that it is well organized on the high value side.

  Steps 11 and 12 described above are repeated as appropriate based on the user's judgment, and further technical considerations are added by the user to finally identify the failure factor.

  According to the present embodiment, the data classification unit (classification means) 4 classifies the basic data group DA00 into a defective second data group DA2 and a non-defective first data group DA1, and an input attribute condition determination unit. 100 creates a regression tree that separates the second data group DA2 of defective products and the first data group DA1 of non-defective products. At this time, the regression tree is sequentially grown only on the other data group side that is the “data group with the smaller ratio of the second data group DA2 of defective products”. For this reason, the structure is simple and clear as compared with the conventional regression tree, and the cause of failure can be identified efficiently and with high accuracy.

  Further, for each branch of the regression tree, the input attribute condition determination unit 100 obtains an input attribute condition that causes a problem event (second data group DA2 of defective products) for all input attributes as well as the branch condition. Of all the input attribute conditions for the number of times, input attribute conditions having a higher degree of defective product separation are narrowed down and determined as the final failure factor (input attribute condition). For this reason, the cause of failure can be clearly identified without omission.

  Further, the input attribute condition list graph creation unit (input attribute condition list creation unit) 20 calculates the defective product separation degree (second data group separation degree) and / or the number of defects (second) calculated by the input attribute condition determination unit 100. An input attribute condition list graph in which input attribute conditions are arranged in order of priority based on the number of corresponding data groups) is created. By referring to the input attribute condition list graph, the user can see and understand a plurality of failure factors and can know the priority order thereof.

  As described above, according to the present embodiment, the present invention “provides a data analysis device that clearly specifies a factor of a predetermined output attribute (such as a factor of product characteristic failure) with a priority order” is provided. Can achieve other objectives.

  In this embodiment, a review graph creation unit (review graph creation means) 21 is provided that creates a review graph by an operation on the screen of the input attribute condition list graph. For this reason, the user can perform operations on the screen while referring to the order (priority order) of the input attribute conditions in the input attribute condition list graph (that is, while navigating the provisional order for reviewing the analysis results). A review graph can be created.

  In addition, in this review graph, the correspondence between the second data group DA2 of defective products and the data group that satisfies the input attribute conditions that cause the failure is visually emphasized. Can be understood. In addition, the conventional regression tree problem that “the evaluation of the result varies depending on the user” can be solved.

  These effects are as follows: (1) The data classification unit (classification means) 4 determines the basic data group DA00, the second data group DA2 of defective products, and the first data group of non-defective products based on the value of the output attribute. (2) The input attribute condition determination unit 100 determines that “the second data group DA2 of defective products is collected in the data group that satisfies the input attribute condition and satisfies the input attribute condition. The input attribute condition is determined so that the non-defective first data group DA1 is gathered into the non-data group ”, that is, (1) the basic data group is divided into two by the output attribute and (2) the basic by the input attribute This is because the point of interest when evaluating the analysis results is clarified by dividing into data groups.

  As described above, according to the present embodiment, it is possible to achieve the object of the present invention “providing a data analysis device that enables intuitive evaluation of analysis results based on a clear index”.

  Furthermore, in the present embodiment, by operation on the screen of the input attribute condition list graph, creation of a multi-faceted review graph by the review graph creation unit (review graph creation means) 21 and a composite factor graph creation unit (combination factor graph creation) Since the creation of the multi-faceted complex factor graph by means 22 can be repeated as appropriate, the analysis result can be easily evaluated while comparing both graphs.

  The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope shown in the claims, and embodiments obtained by appropriately combining technical means disclosed in different embodiments. Is also included in the technical scope of the present invention.

  For example, in the above-described embodiment, the present embodiment has been described by taking as an example the case where the repetitive operations of Step 3 to Step 6 are performed. In this case, since the output of the second input attribute condition determining unit 19 is the same as the output of the first input attribute condition determining unit 12, the branch condition determining unit (branch condition determining means) 14, the data The dividing unit (dividing unit) 15, the end condition determining unit 16, and the second input attribute condition determining unit (second input attribute condition determining unit) 19 may be omitted from the input attribute condition determining unit 100. .

  In the above embodiment, the second data group DA2 is used as a defective product data group, and a data analysis example for extracting the cause of the failure is shown. However, the second data group DA2 is used as a good data group to obtain a good product. It is good also as data analysis which extracts the conditions for this.

  The data analysis apparatus having the above-described configuration includes the review graph creation unit 21 that can output four types of review graphs and the composite factor graph creation unit 22 that can output three types of composite factor graphs. One of these may be omitted. Further, the number of types of graphs that can be output by the review graph creation unit 21 and the composite factor graph creation unit 22 is not particularly limited, and may be only one type. However, it is preferable that a plurality of types of graphs can be output because a graph suitable for the application can be output and a detailed data analysis result can be evaluated by a combination of a plurality of graphs. Further, the graph that can be output by the review graph creation unit 21 and the composite factor graph creation unit 22 is not limited to the illustrated form.

  In the above embodiment, the input attribute condition list graph creation unit 20 creates the input attribute condition list graph. However, instead of the input attribute condition list graph creation unit 20, an input that outputs a list of the plurality of input attribute conditions determined by the input attribute condition determination unit 100 in a form other than a graph, for example, a list form. An attribute condition list creation unit may be used.

  The data analysis method described above can be realized by the computer executing a data analysis program including processes corresponding to S1 to S12 (steps 1 to 12) in FIG. Therefore, in the data analysis apparatus of FIG. 1, the data analysis program includes a computer, a basic data group storage unit 1, a classification condition setting unit 3, a data classification unit 4, a post-classification basic data group storage unit 5, and an analysis data group extraction unit. 6, first input attribute condition determination unit 12, analysis result data storage unit 13, branch condition determination unit 14, data division unit 15, end condition determination unit 16, defective product data calculation unit 17, second input attribute condition determination This can be realized by functioning as the unit 19, the input attribute condition list graph creation unit 20, the review graph creation unit 21, the composite factor graph creation unit 22, and the rank determination unit 23.

  The data analysis program can be stored in a computer-readable recording medium and provided to the user. Thereby, the data analysis program can be easily provided to the computer. The recording medium may be a built-in medium built in the computer main body, or a removable medium configured to be separable from the computer main body. Examples of the built-in medium include ROM; rewritable nonvolatile memory such as flash memory; and hard disk. The removable media includes optical recording media such as CD-ROM and DVD; magneto-optical recording media such as MO; magnetic recording media such as floppy (registered trademark) disks, cassette tapes and removable hard disks; memory cards and the like. Examples of the medium include a rewritable non-volatile memory such as a medium; a medium including a ROM such as a ROM cassette.

  The program may be configured to be executed by CPU access, or after reading the program stored in the recording medium and transferring the read program to the program storage area of the built-in medium, the program on the built-in medium May be executed by CPU access. Further, the program is not limited to a program that is sold in a state of being stored in a computer-readable recording medium, but is sold in a format that is transferred to a user's computer via a communication network such as the Internet. It may be a thing.

It is a block diagram which shows the structure of the data analyzer which concerns on one Embodiment of this invention. It is a graph used for operation | movement description of the data analyzer which concerns on one Embodiment of this invention, and is a trend graph showing transition for every board | substrate of the value of an output attribute. It is a graph used for operation | movement description of the data analyzer which concerns on one Embodiment of this invention, and is a map showing in-plane distribution of an output attribute. It is a flowchart which shows the flow of a process of the data analyzer which concerns on one Embodiment of this invention. It is the figure which expressed the example (Table 5) of the calculation result by the inferior goods data calculating part 17 in the data analyzer which concerns on one Embodiment of this invention with the Venn diagram. It is the figure which expressed the operation | movement (step 3-step 9) of the input attribute condition determination part 100 in the data analysis apparatus which concerns on one Embodiment of this invention in the form of a regression tree. It is an input attribute condition list graph which the input attribute condition list graph preparation part 20 in the data analyzer which concerns on one Embodiment of this invention outputs. FIG. 8 is an input attribute condition list graph obtained by rearranging the graph of FIG. 7 in descending order of defective product separation degree. 8 is an input attribute condition list graph in which the graph of FIG. 7 is rearranged in descending order of the number of defects. FIG. 8 is an input attribute condition list graph in which the graph of FIG. 7 is rearranged in descending order of the score considering both the defective product separation degree and the number of defects. The review graph created by the review graph creation unit 21 and the composite factor graph created by the composite factor graph creation unit 22 in response to operations on the input attribute condition list graphs (FIGS. 7 to 10) will be described. FIG. It is an example of the trend graph which the review graph preparation part 21 in the data analyzer which concerns on one Embodiment of this invention produces. It is an example of the map which the review graph preparation part 21 in the data analyzer which concerns on one Embodiment of this invention produces. It is an example of the scatter diagram which the review graph preparation part 21 in the data analyzer which concerns on one Embodiment of this invention produces. FIG. 14 is an example of a scatter diagram created by the review graph creation unit 21 in the data analysis apparatus according to the embodiment of the present invention. The scatter diagram of FIG. 14 is divided into series under the condition of input attribute 7> 1.15. It is a graph. It is an example of the frequency distribution map which the review graph preparation part 21 in the data analyzer which concerns on one Embodiment of this invention produces. It is an example of the complex factor pareto chart which the complex factor graph creation part 22 in the data analysis device concerning one embodiment of the present invention creates. It is an example of the factor accumulation graph which the composite factor graph preparation part 22 in the data analyzer which concerns on one Embodiment of this invention produces. It is an example of the best case / worst case frequency distribution diagram created by the composite factor graph creation unit 22 in the data analysis apparatus according to the embodiment of the present invention. It is a figure showing the conventional regression tree. A box-whisker graph is represented by a review result of a conventional regression tree analysis. (A) And (b) is the review result of the conventional regression tree analysis, and represents a histogram.

Explanation of symbols

1 Basic data group storage unit 3 Classification condition setting unit 4 Data classification unit (classification means)
5 Basic data group storage after classification 6 Analysis data group extraction unit (analysis data group extraction means)
12 1st input attribute condition determination part (1st input attribute condition determination means)
13 Analysis result data storage unit 14 Branch condition determination unit (branch condition determination means)
15 Data division unit (division means)
16 end condition determining unit 17 defective product data calculating unit 19 second input attribute condition determining unit (second input attribute condition determining means)
20 Input attribute condition list graph creation unit (input attribute condition list creation means)
21 Review graph creation unit (review graph creation means, second basic data group creation means)
22 Composite factor graph creation unit (Compound factor graph creation means)
100 Input attribute condition determining unit (input attribute condition determining means)

Claims (26)

Data for analyzing the causal relationship between the input attribute and the output attribute and outputting the analysis result of the causal relationship with respect to a basic data group that is a set of data composed of a plurality of input attributes and output attributes An analyzer,
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
An input attribute condition, which is an input attribute condition for bisecting the basic data group, is set for each of the plurality of input attributes, so that the first data group and the second data group are respectively combined. An input attribute condition determining means for determining;
Two data groups obtained by obtaining a specific input attribute condition from the plurality of input attribute conditions determined by the input attribute condition determining means, and by dividing the basic data group based on the specific input attribute condition A review graph that visually represents a correspondence relationship between at least one of the first data group and the second data group obtained by bisectioning the basic data group based on the value of the output attribute And a review graph creating means. Data for analyzing the causal relationship between the input attribute and the output attribute and outputting the analysis result of the causal relationship with respect to a basic data group that is a set of data composed of a plurality of input attributes and output attributes An analyzer,
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
An input attribute condition, which is an input attribute condition for bisecting the basic data group, is set for each of the plurality of input attributes, so that the first data group and the second data group are respectively combined. An input attribute condition determining means for determining;
Each of the data composed of the plurality of input attributes and output attributes belongs to each of a plurality of groups, and the input attributes and outputs belonging to the group in units of the plurality of groups. A second basic data group creating means for creating a second basic data group representing a data group obtained by averaging attribute values;
2 obtained by obtaining a specific input attribute condition from among the plurality of input attribute conditions determined by the input attribute condition determining means, and by dividing the second basic data group based on the specific input attribute condition The correspondence relationship between at least one of the two data groups and at least one of the two data groups obtained by the bisection of the second basic data group based on the condition for classifying the basic data group by the classification means is visually shown. And a review graph creating means for outputting a review graph represented by the data analysis device. An input attribute condition list creating means for outputting a list of the plurality of input attribute conditions determined by the input attribute condition determining means;
The review graph creating means includes a selection operation in which a user selects a specific input attribute condition from the plurality of input attribute conditions included in the list output by the input attribute condition list creating means, and a review graph drawing 3. The method according to claim 1, wherein the selected input attribute condition is acquired as the specific input attribute condition in response to performing the instruction, and the review graph is output. 4. Data analysis device. The input attribute condition determined by the input attribute condition determining means is “the second data group is grouped in a data group that satisfies the input attribute condition, and the first data group is included in a data group that does not satisfy the input attribute condition. Is a condition that is determined to
The review graph creating means outputs a review graph that visually represents at least a correspondence relationship between a data group that satisfies the input attribute condition and a data group that satisfies the output attribute condition corresponding to the second data group. The data analysis apparatus according to claim 1, wherein the data analysis apparatus is provided.   In the list of input attribute conditions, the input attribute condition is satisfied in the basic data group for each of the plurality of input attribute conditions determined by the input attribute condition determining means, and the second data group A second data group separation that represents the number of corresponding second data groups that represents the number of corresponding data and the ratio of the number of data that includes the second data group in the data that satisfies the input attribute condition in the basic data group The data analysis apparatus according to claim 4, wherein a degree is indicated together with the input attribute condition. A second value representing the number of data that satisfies the input attribute condition in the basic data group and corresponds to the second data group with respect to the plurality of input attribute conditions determined by the input attribute condition determining means. Using at least one of the number of corresponding data groups and the second data group separation degree indicating the ratio of the number of data corresponding to the second data group in the data satisfying the input attribute condition in the basic data group as an index And rank determination means for ranking according to importance as a factor of the output attribute condition corresponding to the second data group,
6. The data analysis apparatus according to claim 4 or 5, wherein the input attribute condition list creation unit indicates a result of ranking by the rank determination unit in the list of input attribute conditions. In each of the data composed of a plurality of input attributes and output attributes in the basic data group or the second basic data group, a discrimination key is given, respectively.
The review graph creating means includes:
The target data group for which the review graph is to be created is the basic data group or the second basic data group,
For the target data group, a first graph showing a change for each discrimination key of an input attribute corresponding to the acquired specific input attribute condition, and a first graph showing a change for each discrimination key of the output attribute A graph showing both of the graphs of 2 is output as the review graph,
7. The marking according to claim 4, wherein in the first graph, marking processing is applied to data satisfying the input attribute condition and satisfying an output attribute condition corresponding to the second data group. The data analysis apparatus according to claim 1. The basic data group includes a plurality of data corresponding to different time points,
The review graph creating means includes:
The target data group for which the review graph is to be created is the basic data group or the second basic data group,
For the target data group, both a first trend graph showing the time change of the input attribute corresponding to the acquired specific input attribute condition and a second trend graph showing the time change of the output attribute are shown together. The first review graph is output as the review graph,
In the first trend graph, marking processing is performed on data satisfying the input attribute condition and satisfying the output attribute condition corresponding to the second data group. The data analysis device according to any one of claims. The basic data group includes a plurality of data corresponding to different positions in the plane,
The review graph creating means includes:
The target data group for which the review graph is to be created is the basic data group or the second basic data group,
A first map representing the in-plane distribution of the input attribute corresponding to the acquired specific input attribute condition and a second map representing the in-plane distribution of the output attribute are shown together for the target data group. Outputting a second review graph as the review graph,
7. The marking according to claim 4, wherein in the first map, marking processing is performed on data satisfying the input attribute condition and satisfying an output attribute condition corresponding to the second data group. The data analysis apparatus according to claim 1. The review graph creating means includes:
The target data group for which the review graph is to be created is the basic data group or the second basic data group,
For the target data group, a third review graph that is a scatter diagram showing a relationship between the input attribute corresponding to the acquired specific input attribute condition and the output attribute is output as the review graph. In the scatter diagram, marking processing is performed on data satisfying the input attribute condition and satisfying the output attribute condition corresponding to the second data group. The data analysis apparatus according to item 1.   In response to the user instructing the grouping by another input attribute condition different from the input attribute condition corresponding to the axis of the scatter diagram, the review graph creating means The data analysis apparatus according to claim 10, wherein the data that does not satisfy the other input attribute condition is shown in a scatter diagram in a different form. The review graph creating means includes:
The target data group for which the review graph is to be created is the basic data group or the second basic data group,
Separating the target data group into a data group satisfying the acquired specific input attribute condition and a data group not satisfying the input attribute condition;
The data analysis according to any one of claims 4 to 6, wherein a fourth review graph indicating the frequency distribution of the output attribute in each of the data groups is output as the review graph. apparatus. Data for analyzing the causal relationship between the input attribute and the output attribute and outputting the analysis result of the causal relationship with respect to a basic data group that is a set of data composed of a plurality of input attributes and output attributes An analyzer,
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
An input attribute condition, which is an input attribute condition for dividing the basic data group into two parts, is set for each of the plurality of input attributes so that the first data group and the second data group are combined. , Input attribute condition determining means for determining each,
Two or more specific input attribute conditions are acquired from the plurality of input attribute conditions determined by the input attribute condition determining means, and the two or more input attribute conditions are entangled based on the condition A composite representing a correspondence relationship between a data group extracted from a basic data group and at least one of the first data group and the second data group obtained by bisecting the basic data group based on the value of the output attribute A data analysis apparatus comprising: composite factor graph creating means for outputting a factor graph. An input attribute condition list creating means for outputting a list of the plurality of input attribute conditions determined by the input attribute condition determining means;
The composite factor graph creating means includes a selection operation in which a user selects a specific input attribute condition from a plurality of the input attribute conditions included in the list output by the input attribute condition list creating means, and a composite factor In response to performing a graph drawing instruction, the input attribute condition including the selected input attribute condition is acquired as the two or more specific input attribute conditions, and the composite factor graph is output. The data analysis apparatus according to claim 13. The input attribute condition determined by the input attribute condition determining means is “the second data group is grouped in a data group that satisfies the input attribute condition, and the first data group is included in a data group that does not satisfy the input attribute condition. Is a condition that is determined to
In the list of input attribute conditions, the input attribute condition is satisfied in the basic data group for each of the plurality of input attribute conditions determined by the input attribute condition determining means, and the second data group A second data group separation that represents the number of corresponding second data groups that represents the number of corresponding data and the ratio of the number of data that includes the second data group in the data that satisfies the input attribute condition in the basic data group The data analysis apparatus according to claim 14, wherein a degree is indicated together with the input attribute condition. The input attribute condition determined by the input attribute condition determining means is “the second data group is grouped in a data group that satisfies the input attribute condition, and the first data group is included in a data group that does not satisfy the input attribute condition. Is a condition that is determined to
A second value representing the number of data that satisfies the input attribute condition in the basic data group and corresponds to the second data group with respect to the plurality of input attribute conditions determined by the input attribute condition determining means. Using at least one of the number of corresponding data groups and the second data group separation degree indicating the ratio of the number of data corresponding to the second data group in the data satisfying the input attribute condition in the basic data group as an index And rank determination means for ranking according to importance as a factor of the output attribute condition corresponding to the second data group,
16. The data analysis apparatus according to claim 14 or 15, wherein the input attribute condition list creation unit indicates a result of ranking by the rank determination unit in the list of input attribute conditions. The input attribute condition determined by the input attribute condition determining means is “the second data group is grouped in a data group that satisfies the input attribute condition, and the first data group is included in a data group that does not satisfy the input attribute condition. Is a condition that is determined to
The composite factor graph creating means includes:
Two specific input attribute conditions are acquired from the plurality of input attribute conditions determined by the input attribute condition determining means,
In the basic data group, a result of calculating the number of data satisfying both the two input attribute conditions and corresponding to the second data group, and
In the basic data group, the result of calculating the number of data satisfying only one of the two input attribute conditions and corresponding to the second data group is shown as the first data The data analysis apparatus according to any one of claims 13 to 16, wherein a complex factor graph is output as the complex factor graph. The input attribute condition determined by the input attribute condition determining means is “the second data group is grouped in a data group that satisfies the input attribute condition, and the first data group is included in a data group that does not satisfy the input attribute condition. Is a condition that is determined to
The composite factor graph creating means includes:
Obtaining two or more specific input attribute conditions from among the plurality of input attribute conditions determined by the input attribute condition determining means, with ranking,
The second data group corresponding number representing the number of data satisfying the input attribute condition in the basic data group and corresponding to the second data group in the order of the order of the input attribute conditions acquired with the ranking. The data analysis apparatus according to any one of claims 13 to 16, wherein a second composite factor graph indicating a result of sequentially integrating the two is output as the composite factor graph. The input attribute condition determined by the input attribute condition determining means is “the second data group is grouped in a data group that satisfies the input attribute condition, and the first data group is included in a data group that does not satisfy the input attribute condition. Is a condition that is determined to
The composite factor graph creating means includes:
Extract from the basic data group a data group that satisfies all of the acquired two or more input attribute conditions and a data group that satisfies all of the exclusive conditions for each of the acquired two or more input attribute conditions. And
The third composite factor graph indicating the frequency distribution of the output attribute in each of the data groups is output as the composite factor graph. Data analysis device. The input attribute condition determining means includes
Analysis data group extraction means for extracting an analysis data group to be analyzed from the basic data group after classification by the classification means;
An input attribute condition, which is an input attribute condition for dividing the analysis data group into two parts, is set for each of the plurality of input attributes so that the first data group and the second data group are respectively combined. , First input attribute condition determining means for determining each,
Among the input attribute conditions determined by the first input attribute condition determining means, a branch condition determining means for determining, as a branch condition, the input attribute condition having the highest probability of cutting out the second data group;
Dividing means for dividing the analysis data group into two data groups based on the branch conditions determined by the branch condition determining means;
A second input attribute for selecting a highly important input attribute condition as a factor of an output attribute condition corresponding to the second data group from the input attribute conditions determined by the first input attribute condition determining means Including a condition determining means,
The analysis data group extraction unit extracts only a data group having a smaller ratio of the second data group from the two data groups divided by the division unit as a new analysis data group, and the analysis data group A series of processing consisting of processing by the extracting means, processing by the first input attribute condition determining means, processing by the branch condition determining means, and processing by the dividing means is repeatedly executed,
The second input attribute condition determining unit is configured to select, for each of the input attributes, a plurality of input attribute conditions in the input attribute determined by the repetition process of the first input attribute condition determining unit. The input attribute condition having a high degree of importance is selected from the above, and the selected input attribute condition is determined as an output of the input attribute condition determining means. The data analysis device described. Data for analyzing a causal relationship between the input attribute and the output attribute and extracting information indicating the causal relationship with respect to a basic data group that is a set of data composed of a plurality of input attributes and output attributes An analyzer,
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
Analysis data group extraction means for extracting an analysis data group to be analyzed from the basic data group after classification by the classification means;
An input attribute condition, which is an input attribute condition for dividing the analysis data group into two parts, is set for each of the plurality of input attributes so that the first data group and the second data group are respectively combined. , First input attribute condition determining means for determining each,
Among the input attribute conditions determined by the first input attribute condition determining means, a branch condition determining means for determining, as a branch condition, the input attribute condition having the highest probability of cutting out the second data group;
Splitting means for dividing the analysis data group into two based on the branch condition determined by the branch condition determining means,
The analysis data group extraction unit extracts only a data group having a smaller ratio of the second data group from the data group divided by the division unit as a new analysis data group, and the analysis data group extraction unit A series of processes consisting of the process according to the above, the process by the first input attribute condition determining means, the process by the branch condition determining means, and the process by the dividing means are repeatedly executed. Data analysis device. The computer analyzes a causal relationship between the input attribute and the output attribute with respect to a basic data group that is a set of data including a plurality of input attributes and an output attribute, and obtains the analysis result of the causal relationship. A data analysis program for functioning as a data analysis device for output,
Computer
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
An input attribute condition, which is an input attribute condition for bisecting the basic data group, is set for each of the plurality of input attributes, so that the first data group and the second data group are respectively combined. An input attribute condition determining means to determine, and a specific input attribute condition is acquired from the plurality of input attribute conditions determined by the input attribute condition determining means, and the basic data group based on the specific input attribute condition is acquired. Correspondence relationship between at least one of two data groups obtained by bisection and at least one of the first data group and the second data group obtained by bisection of the basic data group based on the value of the output attribute Analysis program for functioning as a review graph creation means for outputting a review graph that visually represents a message. The computer analyzes a causal relationship between the input attribute and the output attribute with respect to a basic data group that is a set of data including a plurality of input attributes and an output attribute, and obtains the analysis result of the causal relationship. A data analysis program for functioning as a data analysis device for output,
Computer
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
An input attribute condition, which is an input attribute condition for bisecting the basic data group, is set for each of the plurality of input attributes, so that the first data group and the second data group are respectively combined. Input attribute condition determining means to determine,
Each of the data composed of the plurality of input attributes and output attributes belongs to each of a plurality of groups, and the input attributes and outputs belonging to the group in units of the plurality of groups. A second basic data group creating means for creating a second basic data group representing a data group in which attribute values are averaged, and the plurality of input attribute conditions determined by the input attribute condition determining means And a condition obtained by classifying the basic data group by the classifying unit and at least one of two data groups obtained by bisection of the second basic data group based on the specific input attribute condition A review graph that outputs a review graph visually representing a correspondence relationship between at least one of two data groups obtained by bisection of the second basic data group based on Data analysis program to function as a view graph creation means. The computer analyzes a causal relationship between the input attribute and the output attribute with respect to a basic data group that is a set of data including a plurality of input attributes and an output attribute, and obtains the analysis result of the causal relationship. A data analysis program for functioning as a data analysis device for output,
Computer
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
An input attribute condition, which is an input attribute condition for dividing the basic data group into two parts, is set for each of the plurality of input attributes so that the first data group and the second data group are combined. , Input attribute condition determining means for determining each, and acquiring two or more specific input attribute conditions from among the plurality of input attribute conditions determined by the input attribute condition determining means, and the two or more specific input attribute conditions The first data group and the second data group obtained by bifurcation of a data group extracted from the basic data group based on a condition entangled with an input attribute condition, and the basic data group based on a value of the output attribute A data analysis program for functioning as a composite factor graph creating means for outputting a composite factor graph representing a correspondence relationship with at least one of the above. In order to analyze a causal relationship between the input attribute and the output attribute and extract information indicating the causal relationship with respect to a basic data group that is a set of data composed of a plurality of input attributes and output attributes Data analysis program
Computer
Classification means for classifying the basic data group into a first data group and a second data group based on the value of the output attribute;
Analysis data group extraction means for extracting an analysis data group to be analyzed from the basic data group after classification by the classification means;
An input attribute condition, which is an input attribute condition for dividing the analysis data group into two parts, is set for each of the plurality of input attributes so that the first data group and the second data group are respectively combined. , First input attribute condition determining means for determining each,
A branch condition determining means for determining, as a branch condition, the input attribute condition having the highest probability of cutting out the second data group among the input attribute conditions determined by the first input attribute condition determining means; Based on the branch condition determined by the branch condition determining means, the analysis data group is made to function as a dividing means for dividing into two parts,
The analysis data group extraction unit extracts only a data group having a smaller ratio of the second data group from the data group divided by the division unit as a new analysis data group, and the analysis data group extraction unit A series of processes consisting of the process according to the above, the process by the first input attribute condition determining means, the process by the branch condition determining means, and the process by the dividing means are repeatedly executed. Data analysis program.   A computer-readable recording medium on which the data analysis program according to any one of claims 22 to 25 is recorded.

Images