JP4347099B2 - Data analysis apparatus, data analysis method, data analysis program, and recording medium - Google Patents

Description

  The present invention provides an output attribute (object attribute) to be analyzed, such as characteristics of a product manufactured in a manufacturing process, and an input attribute (explanatory attribute) that is an attribute affecting the output attribute, such as a manufacturing process condition. The present invention relates to a data analysis apparatus, a data analysis method, and a data analysis program for analyzing a causal relationship with a data.

  A decision tree technique is known as an effective technique for analyzing a causal relationship between an output attribute and an input attribute (see Patent Document 1). In this method, a tree structure is created in which the values of output attributes are well organized in leaf portions that are sequentially cut by the values of input attributes.

  FIG. 11 is an example of a decision tree described in the prior art section of Patent Document 1 (see paragraphs [0002] to [0005] and FIG. 22 of Patent Document 1), and the data group in Table 1 is analyzed. It is targeted. The data group in Table 1 is a set in which twelve pieces of data including a set of four input attribute values x1, x2, x3, and x4 and a value of an output attribute y corresponding to these input attributes are collected. In a decision tree created by this method (hereinafter referred to as “conventional decision tree-1”), as shown in FIG. 11, the values X, Y, and Z of the output attribute y are input attributes x1, x2 , X3, each is well separated. In FIG. 11, a, b, c, and d are values of x2, 1, 2, 3, and 4 are values of x3, and A, B, C, and D are values of x1.

  However, in the conventional decision tree-1 shown in FIG. 11, when data is classified, it is classified into data sets corresponding to the number of values (number of types of attribute values) taken by the input attribute. For example, since the input attribute x2 takes four types of values (a, b, c, d), it is classified into four sets by classification based on the input attribute x2. Therefore, if the number of values that the input attribute takes increases, the decision tree may become complicated.

  As a solution to this problem, Patent Document 1 proposes a decision tree in which attribute values to be grouped are represented by one label for each attribute, and data is classified by the label.

12A to 12D are label hierarchies described in Examples of Patent Document 1 (see paragraphs [0010] to [0028] and FIG. 13 of Patent Document 1). In this embodiment, for example, for the x3 attribute composed of four types of attribute values (1, 2, 3, 4), the x3 attribute values “1” and “2” are labeled “2.5 or less”, and The x3 attribute values “3” and “4” are labeled “2.5 or more” to represent the hierarchical structure. FIG. 13 (paragraphs [0010] to [0028] in FIG. 13 and FIG. 13 shows a decision tree created using this label hierarchical structure (hereinafter, this decision tree will be referred to as a conventional decision tree-2). 14), which is much simpler than the conventional decision tree-1 shown in FIG.
JP-A-8-314725 (Publication date: November 29, 1996) Atsushi Otaki, Yuji Horie, Dan Steinberg, "Applied binary tree analysis method-by CART-", Nikka Giren, July 6, 1998, P44-P47

  The problem of the prior art will be described using the case where the conventional decision tree-2 (FIG. 13) is applied to cause analysis of product characteristic defects in the manufacturing process of a product such as a device.

  Now, the input attributes x1, x2, x3, and x4 in Table 1 are various process data and in-line inspection data in the product manufacturing process, and the output attribute y is the product characteristic data. The output attribute y = Y is the product characteristic. It shall correspond to a defect. Then, it is assumed that the process engineer investigates the cause of the product characteristic defect using the conventional decision tree-2 described in Patent Document 1 for the product characteristic defect y = Y.

  In decision tree-2 (FIG. 13) of Patent Document 1, classification is performed according to the label hierarchy described above, so the decision tree is simple.

However, the condition of product characteristic failure where y = Y is
・ "X2 = c or d"
Or
“X2 = a or b” and “x3 ≧ 2.5” and “x1 = C or D”
It appears in a plurality of places (two places in this example) in the form of a hierarchical structure. For this reason, specifically, it is difficult for the process engineer to determine the cause of the product characteristic failure, such as “Which range of each input attribute is the product characteristic bad?”. For example, for the input attribute x2, it is difficult to understand which is the worse condition, “x2 = c or d” or “x2 = a or b”. Further, the condition “x1 = C or D” is a cause of failure only in the case of a combination with the conditions “x2 = a or b” and “x3 ≧ 2.5”, and the cause of failure in other cases. It is difficult to judge whether or not.

  In the example of FIG. 13, the decision tree is simple because there are only four input attributes and only four types of attribute values. However, when enormous data of the product manufacturing process is used, a product with y = Y is used. Characteristic defects appear everywhere in the decision tree, and branching under the same input attribute under different conditions often occurs many times.

  That is, in the conventional decision tree-2, there is no clear index of “which condition should be extracted as the cause of failure among the branch conditions”, and the process engineer determines the cause of the product characteristic failure. There was a problem that it was difficult.

  Furthermore, as a second problem, in order to create the simple decision tree-2 shown in FIG. 13, it is necessary to predefine the label hierarchical structure shown in FIG. Is not applicable when there is no idea of the attribute values to be summarized. In a manufacturing site of a product such as an actual device, there are process data and in-line inspection data having about 10 to 100 attributes per process, and the values are multi-valued and distributed in a very wide range. Furthermore, due to the influence of disturbance (attributes that cannot be detected as input attributes), the values of output attributes often vary even if the values of the input attributes are the same. Under these circumstances, it is very difficult even for an experienced process engineer to find an attribute value that is collected as one label for each input attribute.

  The present invention has been made in view of the above-described conventional problems, and a first object thereof is to extract a factor of a predetermined output attribute based on a clear index, a data analysis method, a data analysis method, and data To provide analysis programs and recording media. A second object is to provide a data analysis apparatus capable of deriving a causal relationship between an output attribute and an input attribute in a simple form without defining a label hierarchical structure in advance.

  In order to solve the above-described problem, the data analysis apparatus of the present invention provides a basic data group, which is a set of data composed of a plurality of input attributes and output attributes, for the input attributes and the output attributes. A data analysis apparatus for analyzing causal relationships and extracting information indicating the causal relationships, wherein the basic data group is classified into a first data group and a second data group according to the value of the output attribute. Each of each of the analysis data group, a classification means for assigning a classification flag, an 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, and For each of all the input attribute conditions that can be taken by the input attribute, “if the input attribute satisfies the input attribute condition, the data belongs to the second data group in the analysis data group, and the input attribute First evaluation means for calculating an input attribute condition evaluation index representing the probability of the first association rule that the data belongs to the first data group in the analysis data group if the force attribute condition is not satisfied And an input attribute condition determining means for determining an input attribute condition having the maximum input attribute condition evaluation index as an input attribute condition satisfying the first correlation rule for each of the input attributes of the analysis data group And for each of the input attribute conditions determined by the input attribute condition determining means, 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 Among the input attribute conditions determined by the second data group separation degree computing means for calculating the second data group separation degree and the input attribute condition determining means, the basic data An input attribute condition having a second data group separation degree having a value larger than a second data group content ratio that represents a ratio of the number of data pieces of the second data group included in the data group is defined in the second data group. It includes a factor extracting means for extracting as information indicating the factor of the corresponding output attribute condition.

  According to the above invention, the classification means classifies the basic data group into the first data group and the second data group, and the analysis data group extraction means selects the input attribute and output attribute from the classified basic data group. Analytical data group to analyze the causal relationship is extracted. For example, the first data group is a data group having a non-defective product output attribute, and the second data group is a data group having an output attribute representing a problem event such as a defective product.

  For each of all the input attribute conditions that can be taken by each input attribute of the analysis data group, the first evaluation means “if the input attribute satisfies the input attribute condition, it belongs to the second data group in the analysis data group Data, and if the input attribute does not satisfy the input attribute condition, the input attribute condition evaluation index indicating the probability of the first correlation rule is calculated as “data belonging to the first data group in the analysis data group”. The input attribute condition determining means determines the input attribute condition having the maximum input attribute condition evaluation index as the input attribute condition satisfying the first correlation rule for each input attribute of the analysis data group. The second data group separability calculating means calculates a second data group separability for each of the input attribute conditions determined by the input attribute condition determining means.

  The second data group separation degree of each input attribute condition represents 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. Of the input attribute conditions determined by the attribute condition determining means, the second data group having a value larger than the second data group content ratio representing the ratio of the number of data of the second data group included in the basic data group An input attribute condition having a degree of separation is extracted as information indicating a factor of an output attribute condition corresponding to the second data group.

  In this way, the data analysis apparatus analyzes the causal relationship between the input attribute and the output attribute for the basic data group, and extracts information indicating the causal relationship.

  In the above causal analysis, the factor extracting unit corresponds to the second data group from among the input attribute conditions determined by the input attribute condition determining unit based on a clear index of the second data group separation degree. The factor of the output attribute condition, that is, the input attribute condition in which the data becomes the second data group is extracted. Therefore, no matter how complex the decision tree is, it is possible to clearly extract the factor of the output attribute condition corresponding to the second data group. For example, if the output attribute of the second data group represents a problem phenomenon such as a defective product, the second data group separation degree is a defective product separation degree indicating the accuracy of defective product extraction, and how complex the decision tree is. It is possible to clearly understand the cause of the problem phenomenon.

  Further, since the second data group separation degree such as the defective part separation degree is used as the evaluation index, it is possible to prioritize a plurality of factors (input attribute conditions) extracted by the factor extracting means.

  Furthermore, even if the factor extraction means is a condition other than the branch condition in the decision tree, it extracts all conditions with a high degree of second data group separation such as defective product separation, so there are competing factors in the branch condition. Even so, you can catch it without fail.

  As described above, it is possible to provide a data analysis apparatus that extracts a factor of a predetermined output attribute based on a clear index.

  In order to solve the above problem, the data analysis apparatus of the present invention provides, for each of the input attribute conditions determined by the input attribute condition determining means, “if the input attribute satisfies the input attribute condition, the analysis data A second evaluation means for calculating a division rule evaluation value representing the probability of the second association rule that the data is included in the second data group in the group, and the input attribute condition determination means Based on the input attribute condition having the largest division rule evaluation value among the input attribute conditions, the analysis data group is divided into a factor data group that satisfies the input attribute condition and other data that does not satisfy the input attribute condition. Dividing means for dividing the data into groups, wherein the analysis data group extracting means adds at least one of the factor data group and the other data group divided by the dividing means to a new previous group. Extracting as an analysis data group, processing by the analysis data group extraction means, processing by the first evaluation means, processing by the input attribute condition determination means, processing by the second data group separation degree calculation means, the factor extraction means A series of processes including the process according to the above, the process by the second evaluation means, and the process by the dividing means are repeatedly executed.

  According to the above invention, for each of the input attribute conditions determined by the input attribute condition determining means by the second evaluation unit, “if the input attribute satisfies the input attribute condition, the second data group in the analysis data group The division rule evaluation value representing the probability of the second correlation rule that “the data is included in the data” is calculated, and the dividing means factorizes the analysis data group based on the input attribute condition having the maximum division rule evaluation value. The data is divided into a data group and another data group, and the analysis data group extraction means extracts at least one of the factor data group and the other data group as a new analysis data group. Then, the above series of processing is repeatedly executed.

  More detailed factor analysis results can be obtained by such repeated processing.

Further, when the repeated processing is not performed, even if the accuracy of the input attribute condition evaluation index is low due to the influence of disturbance, this problem can be solved by performing the repeated processing.
Furthermore, in a certain input attribute, the output attribute condition factor corresponding to the second data group is repeated even when the two types are “input attribute is below threshold” and “input attribute exceeds threshold”. Both of these factors can be extracted by this process.

  In order to solve the above-mentioned problem, the data analysis apparatus of the present invention has priority over a plurality of the input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting means. A factor determination unit that selects only a condition with a high degree, and the factor determination unit includes a plurality of the input attribute conditions in the same input attribute extracted by the repeated processing of the factor extraction unit. Thus, the input attribute condition that maximizes the second data group separation degree is selected as the condition with the higher priority.

  According to the above invention, for a plurality of input attribute conditions with the same input attribute extracted by repeated processing of the factor extracting means, a condition having a high priority based on a clear index of the second data group separation degree Therefore, the factor of the output attribute condition corresponding to the second data group can be determined with high accuracy while being in a very simple form.

  In addition, the factor determination means determines the output attribute condition corresponding to the second data group from the input attribute conditions extracted by the repetition process of the factor extraction means based on the clear index of the second data group separation degree. The factor, that is, the input attribute condition in which the data becomes the second data group is determined. Therefore, no matter how complicated the decision tree is, it is possible to clearly determine the factor of the output attribute condition corresponding to the second data group. For example, if the output attribute of the second data group represents a problem phenomenon such as a defective product, the second data group separation degree is a defective product separation degree indicating the accuracy of defective product extraction, and how complex the decision tree is. It is possible to clearly understand the cause of the problem phenomenon.

  Further, since the second data group separation degree such as the defective part separation degree is used as an evaluation index, it is possible to prioritize a plurality of factors (input attribute conditions) determined by the factor determining means.

  Furthermore, in the process of the above iterative process, the factor extracting means extracts all conditions having a high second data group separation degree such as a defective product separation degree, even if the condition is a condition other than the branch condition in the decision tree. The factor determining means determines the factor of the output attribute condition corresponding to the second data group based on a clear index of the second data group separation degree from the input attribute conditions extracted by the repeated processing of the factor extracting means. That is, since the input attribute condition for which the data becomes the second data group is determined, even if a competing factor exists in the branch condition, the output attribute condition corresponding to the second data group can be surely met without missing the factor. The factors can be determined.

  In order to solve the above-mentioned problem, the data analysis apparatus of the present invention has priority over a plurality of the input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting means. A plurality of input attribute conditions for the same input attribute extracted by the repetition process of the factor extraction means, wherein the input attribute condition is a threshold value. In the case where the factor determination means is divided into a second pattern that is “the input attribute exceeds a threshold value” and the second pattern that is “the input attribute exceeds the threshold value”, the factor determination means includes the second pattern in the first pattern. One input attribute condition that maximizes the data group separation degree, and one input attribute condition that maximizes the second data group separation degree in the second pattern, It is characterized in that ahead of is selected as high condition.

  According to the above invention, the plurality of input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting means are “input attribute is below threshold” and “input attribute is threshold Even in the case of the two patterns “exceeding”, it is possible to determine the factor of the output attribute condition corresponding to the second data group with high accuracy in a very simple form.

  In order to solve the above problems, the data analysis apparatus according to the present invention, the analysis data group extraction unit extracts only the other data group from the data group divided by the division unit as a new analysis data group. It is characterized by that.

  According to the above invention, since only the other data group among the data groups divided by the dividing means is subjected to the above-described repetitive processing as a new analysis data group, the output attribute condition corresponding to the second data group A simple and sufficient factor analysis result is obtained for the factor analysis.

  In addition, since the other data group is processed as a new analysis data group, the analysis can be performed by excluding the influence of the input attribute condition that satisfies the second correlation rule in the previous iteration process, Thereby, a new factor of the output attribute condition corresponding to the second data group can be extracted with high accuracy.

  In order to solve the above-described problem, the data analysis apparatus of the present invention further includes an end condition determining unit that determines whether or not an end condition is satisfied, and the end condition determining unit determines that the end condition is satisfied. Then, the execution of the series of processes is ended, and the end condition determining means is configured such that the number of data in the second data group in the analysis data group extracted by the analysis data group extraction means is 0. It is characterized in that it is determined as the end condition.

  According to the above invention, the detailed process of the factor analysis is obtained because the process is repeated until the number of data in the second data group in the analysis data group becomes zero.

  In the data analysis apparatus of the present invention, in order to solve the above-described problem, all the input attributes in the analysis data group are numerical attributes, and the first evaluation unit is configured for all the numerical values of the input attributes. In the first data group of the analysis data group, the ratio of the number of data whose input attribute is equal to or less than the numerical value is calculated as a first frequency, and in the second data group of the analysis data group, Frequency calculation means for calculating the ratio of the number of data whose input attribute is less than or equal to the value as a second frequency, and the difference between the first frequency and the second frequency for all the numerical values of each input attribute And a difference calculating means.

According to the above invention, the difference between the first frequency and the second frequency calculated by the difference calculation means for each numerical value of the input attribute is expressed as “if the input attribute is less than or equal to the numerical value, it belongs to the second data group. If the input attribute exceeds the numeric value, the data belongs to the first data group. Or, “If the input attribute exceeds the numeric value, the data belongs to the second data group. It can be used as an input attribute condition evaluation index representing the certainty of the first association rule that “is the data belonging to the first data group if is less than or equal to that value”.

  Thus, since the difference between the first frequency and the second frequency is used as an input attribute condition evaluation index when the first data group and the second data group are separated, the label hierarchical structure is defined in advance. It is possible to determine an input attribute condition that satisfies the first correlation rule without any trouble and with simple processing. Accordingly, it is possible to provide a data analysis apparatus that can derive a causal relationship between an output attribute and an input attribute in a simple manner without defining a label hierarchical structure in advance.

  In the data analysis apparatus of the present invention, in order to solve the above-described problem, the second evaluation unit performs the first analysis of the analysis data group for each of the input attribute conditions determined by the input attribute condition determination unit. The ratio of the ratio of the number of data satisfying the input attribute condition in the second data group of the analysis data group to the ratio of the number of data satisfying the input attribute condition in the data group is calculated as the division rule evaluation value. It is characterized by being.

  According to the above invention, with respect to each of the input attribute conditions determined by the input attribute condition determining means, the ratio that can be detected from the first data group and separated from the first data group by the input attribute condition is used as the division rule evaluation value. It can be calculated.

  In order to solve the above-described problem, the data analysis apparatus of the present invention further includes a classification condition setting unit that sets a classification condition, and the classification unit is configured based on a comparison between the value of the output attribute and the classification condition. It is characterized by classifying basic data groups.

  According to the above invention, the basic data group can be classified into the first data group and the second data group based on predetermined classification conditions set as appropriate. That is, an output attribute condition corresponding to the second data group can be set as appropriate, and the factor can be extracted or determined.

  In order to solve the above problems, the data analysis method of the present invention uses the data analysis device to analyze the causal relationship with respect to the basic data group and extract information indicating the causal relationship. A classification step of classifying the basic data group into the first data group and the second data group according to a value of the output attribute and assigning the classification flag by the classification means; An analysis data group extraction step of extracting the analysis data group from the basic data group after classification by the classification means by the analysis data group extraction means; and the analysis data group by the first evaluation means. A first evaluation step of calculating the input attribute condition evaluation index for each of all the input attribute conditions that can be taken by each of the input attributes, and the input attribute condition determining means An input attribute condition determining step for determining the input attribute condition having the maximum input attribute condition evaluation index as the input attribute condition satisfying the first correlation rule for each of the input attributes of the analysis data group A second data group separation degree computing step for computing the second data group separation degree for each of the input attribute conditions determined by the input attribute condition determining means by the second data group separation degree computing means And the input having the second data group separation degree having a value larger than the second data group content rate in the input attribute conditions determined by the input attribute condition determining means by the factor extracting means A factor extracting step of extracting the attribute condition as information indicating a factor of the output attribute condition corresponding to the second data group.

  According to the above invention, the classification means classifies the basic data group into the first data group and the second data group in the classification step, and the analysis data group extraction means performs the basic data group after the classification in the analysis data group extraction step. An analysis data group to be analyzed for the causal relationship between the input attribute and the output attribute is extracted. For example, the first data group is a data group having a non-defective product output attribute, and the second data group is a data group having an output attribute representing a problem event such as a defective product.

  In the first evaluation step, the first evaluation means, for each of all the input attribute conditions that can be taken by each input attribute of the analysis data group, “if the input attribute satisfies the input attribute condition, An input indicating the probability of the first correlation rule that the data belongs to the second data group and if the input attribute does not satisfy the input attribute condition, the data belongs to the first data group in the analysis data group. The attribute condition evaluation index is calculated, and in the input attribute condition determination step, the input attribute condition determination means determines the first input attribute condition having the maximum input attribute condition evaluation index for each input attribute of the analysis data group as a first. It is determined as an input attribute condition that satisfies the correlation rule. In the second data group separation degree calculation step, the second data group separation degree calculation means calculates a second data group separation degree for each of the input attribute conditions determined by the input attribute condition determination means.

  The second data group separation degree of each input attribute condition represents 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. The means has a value larger than the second data group content ratio representing the ratio of the number of data of the second data group included in the basic data group in the input attribute condition determined in the input attribute condition determining step. The input attribute condition having the second data group separation degree is extracted as information indicating the cause of the output attribute condition corresponding to the second data group.

  In this way, the data analysis method analyzes the causal relationship between the input attribute and the output attribute for the basic data group, and extracts information indicating the causal relationship.

  In the above causal analysis, the factor extraction step corresponds to the second data group from among the input attribute conditions determined in the input attribute condition determination step based on a clear index of the second data group separation degree. The factor of the output attribute condition, that is, the input attribute condition in which the data becomes the second data group is extracted. Therefore, no matter how complicated the decision tree is, it is possible to clearly grasp the cause of the output attribute condition corresponding to the second data group. For example, if the output attribute of the second data group represents a problem phenomenon such as a defective product, the second data group separation degree is a defective product separation degree indicating the accuracy of defective product extraction, and how complex the decision tree is. It is possible to clearly understand the cause of the problem phenomenon.

  Further, since the second data group separation degree such as the defective article separation degree is used as an evaluation index, it is possible to prioritize a plurality of factors (input attribute conditions) extracted in the factor extraction step.

  Furthermore, in the factor extraction step, even if the condition is other than the branch condition in the decision tree, all the conditions having a high second data group separation degree such as a defective product separation degree are extracted, so there are competing factors in the branch condition. Even so, you can catch it without fail.

  As described above, a data analysis method for extracting a factor of a predetermined output attribute based on a clear index can be provided.

  In the data analysis method of the present invention, in order to solve the above problem, the input attribute is a manufacturing process condition and / or in-line inspection result in a product manufacturing process, and the output attribute is a product quality determination result. The second data group is characterized in that the quality judgment result is a defective data group.

  According to the above invention, it is possible to extract or determine the cause of the defect in the manufacturing process, and it is possible to easily improve the manufacturing process such as an improvement in yield.

  In order to solve the above problems, the data analysis program of the present invention uses a computer, a basic data group that is a set of data composed of a plurality of input attributes and output attributes, depending on the value of the output attribute. An analysis for classifying into a first data group and a second data group and assigning a classification flag and extracting an analysis data group to be analyzed from the basic data group after classification by the classification means For each of the input attribute conditions that can be taken by the data group extraction means and each input attribute of each of the analysis data groups, “if the input attribute satisfies the input attribute condition, the second data in the analysis data group Data belonging to a data group, and if the input attribute does not satisfy the input attribute condition, the data belongs to the first data group in the analysis data group. " A first evaluation means for calculating an input attribute condition evaluation index that represents the complexity, and an input attribute condition having the maximum input attribute condition evaluation index for each of the input attributes of the analysis data group, An input attribute condition determining unit that determines an input attribute condition that satisfies the first correlation rule, and for each of the input attribute conditions determined by the input attribute condition determining unit, the input attribute condition is satisfied in the basic data group The second data group separation degree calculating means for calculating a second data group separation degree, which represents a ratio of the number of data in which the second data group is included in the data, and the input attribute condition determining means determined by the input attribute condition determining means Among the input attribute conditions, the second data group component having a value larger than the second data group content ratio indicating the ratio of the number of data of the second data group included in the basic data group. An input attribute condition with degrees, extracted as the information indicating the cause of the output attribute condition corresponding to the second data group, to function as a factor extraction unit.

  According to the above invention, data analysis for extracting a factor of a predetermined output attribute based on a clear index can be realized by executing a program on a computer.

  In order to solve the above problems, a recording medium of the present invention is a computer-readable recording medium on which the data analysis program is recorded.

  According to the above invention, the data analysis program can be easily provided to a computer.

  As described above, the data analysis apparatus according to the present invention includes a classification unit, an analysis data group extraction unit, a first evaluation unit, an input attribute condition determination unit, a second data group separation degree calculation unit, and a factor extraction. Means. The data analysis method of the present invention includes a classification step, an analysis data group extraction step, a first evaluation step, an input attribute condition determination step, a second data group separation degree calculation step, and a factor extraction step. Contains.

  Therefore, there is an effect that it is possible to provide a data analysis device that extracts a factor of a predetermined output attribute based on a clear index.

  One embodiment of the present invention will be described below.

  First, 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 character-numerical data conversion unit 2, a classification condition setting unit (classification condition setting unit) 3, a data classification unit (classification unit) 4, and after classification. Basic data group storage unit 5, analysis data group extraction unit (analysis data group extraction unit) 6, data row separation unit 7, data string extraction unit 8, frequency calculation unit (frequency calculation unit) 9, frequency cumulative difference calculation unit (difference) (Calculation means) 10, input attribute condition determination unit (input attribute condition determination means) 11, defective product separation degree calculation part (second data group separation degree calculation means) 12, factor extraction part (factor extraction means) 13, frequency accumulation ratio Arithmetic unit (second evaluation unit) 14, data dividing unit (dividing unit) 15, end condition determining unit 16, factor determining unit (factor determining unit) 17, composite factor defect number calculating unit 18, numerical value-character data converting unit 19, analysis result data storage unit 20, And a reserve power unit 21. The data row separation unit 7, the data string extraction unit 8, the frequency calculation unit (frequency calculation unit) 9, and the frequency cumulative difference calculation unit (difference calculation unit) 10 include the first evaluation in the claims. Means.

  Next, the data analysis apparatus and the data analysis method of this 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 is composed of 12 pieces of data having ids (identifiers) of 1 to 12. In Table 1, x1, x2, x3, and x4 are input attributes. The input attribute x1 is a character attribute that takes one of four characters A, B, C, and D. The input attribute x2 is a character attribute that takes one of the four characters a, b, c, and d. The input attribute x3 is a discrete attribute that takes one of four discrete values 1, 2, 3, and 4. The input attribute x4 is a discrete attribute taking any one of four discrete values 10, 20, 30, and 40. The input attribute may be any of a character attribute, a discrete numerical attribute, and a continuous numerical attribute.

  In Table 1, y is an output attribute. The output attribute may be any of a character attribute, a discrete numerical attribute, and a continuous numerical attribute. Here, the output attribute is a character attribute that takes one of the three characters X, Y, and Z.

  The data analysis apparatus and the data analysis method according to the present embodiment analyze a causal relationship between an input attribute and an output attribute, and extract information indicating the causal relationship. Here, the case where y = Y is considered as a problem event, and the cause of the output attribute y being Y is analyzed.

  As an example of the basic data group DA, for example, the input attribute is the manufacturing process condition and / or in-line inspection result (inspection result during the manufacturing line) in the product manufacturing process, the output attribute is the product quality determination result, Data in which the problem event y = Y is a bad quality determination result is exemplified. In this case, by analyzing the causal relationship between the input attribute and the output attribute by the data analysis apparatus and the data analysis method according to the present embodiment and deriving the cause of the problem event y = Y, the cause of the defect in the manufacturing process is extracted or Therefore, it is possible to easily take measures to eliminate the occurrence of defective products. Therefore, it is possible to easily improve the manufacturing process such as improvement in yield.

  As a more specific example of the basic data group DA, for example, input attributes x1, x2, x3, and x4 are process data such as gas flow rate, gas pressure, input power, and film formation time of plasma CVD process, and output. Data in which the attribute y is the film thickness of the thin film to be formed can be mentioned. The values of the input attribute and the output attribute may be any of a continuous numerical attribute, a discrete numerical attribute, and a character attribute. In the case of the character attribute, for example, the output attribute is an example of the film thickness, and is expressed as “large”, “medium”, and “small”.

Hereinafter, the operation of the data analysis apparatus of FIG. 1 will be described using the flowchart of the data analysis method shown in FIG.
[Step 0]
First, the character-numeric data conversion unit 2 performs numeric conversion processing on non-numeric data in the basic data group DA in Table 1 stored in the basic data group storage unit 1 which is a storage means such as a hard disk. (Hereinafter, the data of each table is stored in an appropriate storage means such as a hard disk or RAM, and is read out from these data and is subject to calculation by calculation means such as a CPU). Here, the character attributes in the basic data group DA are converted into numerical attributes (numerical data) according to the following conversion rules (S0). Note that this processing is omitted when the input attribute and output attribute of the basic data group DA are originally numeric attributes.
(X1) A → 1, B → 2, C → 3, D → 4
(X2) a → 1, b → 2, c → 3, d → 4
(X3) No conversion (x4) No conversion (y) X → 1, Y → 2, Z → 3
Through the above processing, each data is converted into numerical data. Then, the character-numeric data conversion unit 2 sends the converted data group DA0 to the data classification unit 4.

  Here, it is preferable that the above conversion rule is set so that the numerical value of the output attribute increases as the input attribute value after conversion becomes larger as much as possible, or vice versa. There is no limitation to the above example. The data group DA0 converted into numerical data by the conversion rule is as shown in Table 2.

The data group DA0 obtained by this conversion is a set of data composed of a plurality of input attributes (description attributes) and output attributes (target attributes), each consisting of a numerical attribute. Hereinafter, the data group DA0 is also referred to as a basic data group.
[Step 1]
The classification condition setting unit 3 performs basic data group DA in accordance with predetermined setting information or in response to a user inputting a problem event attribute value y = Y from an input unit such as a keyboard or a mouse (not shown). A classification condition that is a condition for determining whether or not the output attribute y of the basic data group DA0 corresponding to the problem event y = Y is set and output to the data classification unit 4 (S1). In this example, the condition of the output attribute y of the basic data group DA0 corresponding to the problem event y = Y of the basic data group DA is y = 2.
[Step 2]
Next, based on the comparison between the value of the output attribute y of the basic data group DA0 and the classification condition (the following comparison logic (1) (2)) output from the classification condition setting unit 3, the data classification unit 4 The basic data group DA0 is classified into a first data group DA1 and a second data group DA2, and a classification flag corresponding to each data group is assigned as shown in Table 3 (S2). Hereinafter, the data group in Table 3 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) y ≠ 2 → DA1
(2) y = 2 → DA2

  Here, the second data group DA2 is a data group representing a problem event (for example, a device characteristic failure or the like). That is, the second data group DA2 is a data group in which the output attribute y is an attribute value (y = 2) representing a problem event, and the first data group DA1 is an attribute value (y) in which the output attribute y does not represent a problem event. = 1 or 3).

  The classification by the data classification unit 4 is not limited to the above logic, and may be performed by the logic shown in the following classification condition based on the comparison with the threshold value yth.

(1 ') y> yth → DA1
(2 ′) y ≦ yth → 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 <y ≦ yth2 → DA1 (yth1, yth2: threshold)
(2 ") y≤yth1 OR y> yth2 → DA2
Furthermore, there are a plurality of output attributes (y1, y2), and the conditions expressed as comparison results with individual threshold values for each output attribute are as shown in the following classification conditions based on the logical sum or logical product. You may classify by logic.

(1 ″ ′) y1 ≦ yth1 OR y2> yth2 → DA1
(2 ″ ′) y1> yth1 AND y2 ≦ yth2 → DA2
[Step 3]
The analysis data group extraction unit 6 extracts the analysis data group DA00 ′ to be analyzed from the classified basic data group DA00 and sends it to the data row separation unit 7.

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]
Based on the classification flags of the first data group DA1 and the second data group DA2 in the analysis data group DA00 ′ (in the first processing, the basic data group DA00 after classification: Table 3). Then, the analysis data group DA00 ′ is divided into two, and each data group is extracted. Table 4 shows the first data group DA1 output from the data row separation unit 7, and Table 5 shows the second data group DA2.

In the following description, 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 referred to as a defective product (NG product) data group.
[Step 5]
Next, the data string extraction unit 8 extracts each data string of the input attribute xj (1 ≦ j ≦ 4) from the good product data group DA1 (Table 4) (S5). This data string is called a 1-xj data group.

  Similarly, the data string extraction unit 8 extracts each data string of the input attribute xj (1 ≦ j ≦ 4) from the defective product data group DA2 (Table 5) (S5). This data string is called a 2-xj data group.

  The 1-xj data group is shown in Tables 6 to 9, and the 2-xj data group is shown in Tables 10 to 13.

[Step 6]
The frequency calculation unit 9 inputs each of the 1-xj data group extracted from the non-defective product data group DA1 in step 5 and each of the 2-xj data group extracted from the defective product data group DA2 in step 5 to the input attribute xj. The values are sorted in ascending order by the value of (sorting process 1). Then, for each numerical value of the input attribute xj, 1-xj frequency cumulative% representing the ratio of the number of data less than that value in the first data group and 2 representing the ratio of the number of data less than that value in the second data group. -Xj Frequency cumulative% is calculated (S6).

  Here, Tables 14 to 17 in which Tables 6 to 9 are rearranged in ascending order by the value of the input attribute xj are used, and the number of data in each row (id) in the table is equal to or greater than the position of the data The ratio of the first data group to the total number of data (= 8) is calculated as 1-xj frequency cumulative%. For example, in Table 14, eight x1 values are arranged from top to bottom in order from 1 to 4, and when there are a plurality of the same values of x1, they are arranged from top to bottom in the order of id. Focusing on the 4th x1 = 1 (id = 12) from the beginning, since there are four data (all x1 = 1 in this case) including this value, the 1-x1 frequency cumulative% is 4/8 = 50%. Similarly, using Table 18 to Table 21 in which Table 10 to Table 13 are rearranged in ascending order by the value of the input attribute xj, the number of data in each row (id) is equal to or greater than the position of the data in the table. The ratio of the second data group to the total number of data (= 4) is calculated as 2-xj frequency cumulative%.

  The values of 1-xj frequency cumulative% and 2-xj frequency cumulative% calculated here are shown in Tables 14 to 21.

  Further, the frequency calculation unit 9 is a table of 1-xj data groups that are non-defective product data groups for which 1-xj frequency cumulative% is calculated, and 2-items that are defective product data groups for which 2-xj frequency cumulative% is calculated. The table of the xj data group is joined (joining process). Specifically, for the input attribute x1, Table 14 and Table 18 are combined to combine the x1 frequency accumulation table of Table 22 (a), and for the input attribute x2, Table 15 and Table 19 are combined to The x2 frequency accumulation table is combined with Table 16 and Table 20 for the input attribute x3 to combine the x3 frequency accumulation table of Table 24, and the input attribute x4 is combined with Table 17 and Table 21 to generate the x4 frequency of Table 25. Each accumulation table is created (S6).

  Furthermore, the frequency calculation unit 9 sorts the frequency accumulation tables of Table 22 (a), Table 23, Table 24, and Table 25 in ascending order by the value of the input attribute xj (sorting process 2). Then, after performing the rearrangement process 2, the value immediately above is substituted into the blanks of 1-xj frequency accumulation% and 2-xj frequency accumulation% (substitution process). Thereafter, only the data of the last line among the lines with the same value in the input attribute xj is adopted (duplicate processing). As a result, Tables 26 to 29 are created. For example, when creating the table 26 from the table 22 (a), if the sorting process 2 is performed in the table 22 (a), each value of the 1-x1 frequency accumulation% column and the 2-x1 frequency accumulation% column Each value is arranged with a blank space between each value while maintaining the respective order (Table 22 (b)). In the next substitution process, these blanks are processed (colored portion in Table 22 (c)). For example, since there is no value corresponding to x1 = 2 in the column of 2-x1 frequency accumulation%, x1 = 1 immediately above the column corresponding to x1 = 2 of 2-x1 frequency accumulation% is obtained by substitution processing. The value 25% of the 2-x1 frequency accumulation% corresponding to is substituted. In this Table 22 (c), there are 5 x1, 2 is 2, 3 is 2, 4 is 3, and the final value for each x1 value is obtained by duplication processing. Adopt the row data. For example, the value 50% corresponding to x1 = 1 of the last row is adopted as the value of 1-x1 frequency cumulative% corresponding to x1 = 1, and the value of 1-x1 frequency cumulative% corresponding to x1 = 2 is adopted as the final row. A value of 75% corresponding to x1 = 2 is adopted. Further, the value 100% corresponding to x1 = 4 in the last row is adopted as the value of 2-x1 frequency cumulative% corresponding to x1 = 4.

  Thus, as shown in Tables 26 to 29, for each value of the input attribute xj, the 1-xj frequency representing the ratio of the number of data whose input attribute is less than or equal to the numerical value in the first data group which is a non-defective data group Cumulative% (A; first frequency) and 2-xj frequency cumulative% (B; second frequency) representing the ratio of the number of data whose input attributes are less than or equal to the numerical value in the second data group that is a defective product data group ) Are calculated (S6).

In the above steps 4 to 6, the xj frequency accumulation table of Table 26 to Table 29 is created in order to create the data row separation process (Tables 4 and 5) → the data string extraction process (Tables 6 to 13) → the arrangement. Replacement process 1 → Calculation process of 1-xj frequency accumulation% and 2-xj frequency accumulation% (Table 14 to Table 21) → Combination process (Table 22 (a), Table 23, Table 24, Table 25) → Reordering process 2 → Substitution process → Duplicate process (Tables 26 to 29) has been performed, but the xj frequency accumulation tables of Table 26 to Table 29 are created directly and collectively without performing these individual processes. As such, it may be calculated. Further, only some of the individual processes may be processed at once.
[Step 7]
Next, the frequency accumulation difference calculation unit 10 subtracts the non-defective 1-xj frequency accumulation% (A) from the defective 2-xj frequency accumulation% (B) for each value of the input attribute xj. Xj frequency cumulative difference (= B−A), and second xj frequency cumulative difference (= B−A) obtained by subtracting 2-xj frequency cumulative% (B) of defective products from 1−xj frequency cumulative percent (A) of non-defective products A difference between A and B, that is, (A−B) is calculated (S7). The first xj frequency cumulative difference (= B−A) and the second xj frequency cumulative difference (= A−B) are collectively referred to as an xj frequency cumulative difference. This xj frequency cumulative difference corresponds to the input attribute condition evaluation index in the claims.

  Tables 30 to 33 show the calculation results of the xj frequency cumulative difference (first xj frequency cumulative difference (= BA), second xj frequency cumulative difference (= AB)). Further, the value of the input attribute xj, the non-defective 1-xj frequency cumulative% (A), the defective 2-ch frequency cumulative% (B), the first xj frequency cumulative difference (= B−A), the second FIG. 3 to FIG. 6 show the relationship with the xj frequency cumulative difference (= A−B).

  The xj frequency cumulative difference (first xj frequency cumulative difference (= B−A), second xj frequency cumulative difference (= A−B)) in each numerical value of the input attribute xj indicates that the input attribute xj is less than or equal to the numerical value. This is an index indicating whether the first non-defective product data group DA1 and the second non-defective product data group DA2 are well separated by being divided into two ranges of the input attribute xj exceeding the numerical value.

  That is, the first xj frequency cumulative difference (= B−A) in each numerical value of the input attribute xj is “if the input attribute xj is equal to or smaller than the numerical value, it is data belonging to the second data group DA2 of defective products and input. It represents the probability of the correlation rule that if the attribute xj exceeds the numerical value, the data belongs to the first non-defective data group DA1.

  Further, the second xj frequency cumulative difference (= A−B) in each numerical value of the input attribute xj is “data belonging to the second data group DA2 of defective products if the input attribute xj exceeds the numerical value, If the input attribute xj is equal to or less than the numerical value, it indicates the likelihood of the correlation rule that it is data belonging to the first non-defective data group DA1.

  Each of the two correlation rules is “if the input attribute satisfies the input attribute condition, the data belongs to the second data group DA2 in the analysis data group DA00 ′, and if the input attribute does not satisfy the input attribute condition, The first association rule is “data belonging to the first data group DA1 in the analysis data group DA00 ′”.

The processes in steps 4 to 7 correspond to the first evaluation step in the claims. In addition, the data row separation unit 7, the data string extraction unit 8, the frequency calculation unit (frequency calculation unit) 9, and the frequency cumulative difference calculation unit (difference calculation unit) 10 used for the processes in steps 4 to 7 are patented. It constitutes the first evaluation means in the claims.
[Step 8]
The input attribute condition determination unit 11 determines, for each input attribute xj, among the xj frequency cumulative difference (first xj frequency cumulative difference or second xj frequency cumulative difference) corresponding to each value of xj. A condition that maximizes the value (the value of the input attribute xj and the type of xj frequency cumulative difference (first xj frequency cumulative difference or second xj frequency cumulative difference)) is extracted (S8). In Tables 30 to 33, the extracted conditions are colored.

  The value of the extracted input attribute xj will be referred to as an input attribute threshold value xj-th. As can be seen with reference to FIGS. 3 to 6, the input attribute threshold value xj-th is divided into a range of xj ≦ xj-th and a range of xj> xj-th, so that the non-defective first data group DA1. And the value of the input attribute xj that makes it easy to separate the defective product from the second data group DA2.

  When the first xj frequency cumulative difference (= B−A) is extracted as the type of xj frequency cumulative difference, the input attribute condition “xj ≦ xj−th” is included in the second data group DA2 of defective products. Correspondingly, when the second xj frequency cumulative difference (= A−B) is extracted, the input attribute condition “xj> xj−th” corresponds to the defective second data group DA2.

  Table 34 shows the input attribute conditions determined by the input attribute condition determination unit 11 as conditions corresponding to the second data group DA2 of defective products for each input attribute xj. As an example, the input attribute condition “x2> 2” is determined for the input attribute x2, but this condition is separated from the non-defective first data group DA1 and the second data group DA2 of defective products is high. The conditions that can be detected with accuracy are shown. Also, “x2 ≦ 2”, which is an exclusive condition for the determined input attribute condition “x2> 2”, separates the defective second data group DA2 from the defective first data group DA1 with high accuracy. The conditions that can be detected are shown. These can be better understood with reference to FIG.

  As described above, the input attribute condition determination unit 11 determines the difference (xj frequency accumulation) between the first frequency (1-xj frequency accumulation% of non-defective products) and the second frequency (2-xj frequency accumulation% of defective products). Difference) is used as an input attribute condition evaluation index when separating the first data group of non-defective products and the second data group of defective products, so that the label hierarchy structure is not defined in advance and can be performed with simple processing. The input attribute condition satisfying the first correlation rule can be determined.

In the above description, the processing from step 5 to step 8 is collectively performed for a plurality of input attributes. However, the value of j may be sequentially increased from 1 to 4 and the processing from step 5 to step 8 may be repeated. . Here, although the xj frequency cumulative difference is used as the input attribute condition evaluation index, an index for evaluating the degree of data bias, for example, the improvement degree of the Gini dispersion index described in Non-Patent Document 1, the information gain ( The input attribute condition may be determined using (Gain) or the like.
[Step 9]
For each of the input attribute conditions (Table 34) determined by the input attribute condition determination unit 11, the defective product separation degree calculation unit 12 uses 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 35) and the number of data satisfying the input attribute condition and corresponding to the defective second data group DA2 ("DA2" column in Table 35) ). Then, the defective product separation degree is calculated by dividing the value in the “DA2” column of Table 35 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 is included in the data satisfying the input attribute condition, and the probability of defective product extraction by the input attribute condition (after classification) This represents a defect rate when data belonging to the input attribute condition in the basic data group DA00 is a population, and corresponds to the second data group separation degree in the claims.

  Table 35 shows the total number of data in the basic data group DA00 after classification (= 12: “DA1 + DA2” column), defective products, in the “Total” row, together with the calculation result by the defective product separation degree calculation unit 12. Is a table showing the number of second data groups DA2 (= 4: “DA2” column) and the defective product content rate (= 4/12 = 0.333: “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 is a defective rate when all data in the basic data group DA00 after classification is a population. This corresponds to the content ratio of the second data group in the claims.

The meaning of each column in Table 35 can be easily understood with reference to FIG.
[Step 10]
The factor extraction unit 13 is more than the defective product content rate of the basic data group DA00 after classification (= 0.333: “Defective Product Separation” column in the “Total” row) in each input attribute condition in Table 35. An input attribute condition having a large value of defective product separation is extracted as information indicating a factor of the second data group DA2 of defective products, that is, information indicating a factor of an output attribute condition corresponding to the defective product. Then, this result is stored in the analysis result data storage unit 20.

  In the example of Table 35, all the input attribute conditions x1 to x4 have a defective product separation degree higher than the defective product content rate of the basic data group DA00 after classification, and all the conditions are extracted (Table 36). Each input attribute condition in Table 36 is a condition that includes the second data group DA2 of defective products at a higher rate than the sample randomly selected from the basic data group DA00 after classification, and the second data group of defective products. The cause of the output attribute condition corresponding to the data group DA2 is shown.

  As described above, the input attribute conditions “x1> 2,” “x2> 2,” “x3> 2,” “x4 ≦ 10” are caused as the cause of the problem event (defective product second data group DA2). Extracted.

In steps 1 to 10, the cause of the problem event (defective product second data group DA2) could be extracted. However, the input attribute condition evaluation index (xj frequency cumulative difference) for each input attribute calculated in the process (steps 6 to 7) includes the influence of other input attributes as disturbances. There is a possibility that the accuracy of the input attribute condition evaluation index (xj frequency cumulative difference) has been lowered. Further, in the case of a certain input attribute xj, if the cause of the problem event is of two types “xj ≦ xj−th1” and “xj> xj−th2”, only one of them is extracted. In order to eliminate these points, it is preferable to further perform processing according to the following steps.
[Step 11]
For each input attribute threshold value xj-th (see Tables 30 to 33) extracted by the input attribute condition determination unit 11 (step 8), the frequency accumulation ratio calculation unit 14 performs the non-defective 1-xj frequency accumulation% (A). , A ratio of 2-xj frequency cumulative% (B) of defective products (= B / A: hereinafter referred to as frequency cumulative lower ratio), or a value obtained by subtracting 1-xj frequency cumulative% (A) of non-defective products from 100 Ratio (= (100−B) / (100−A)) (= (100−B) / (100−A)): (Referred to as cumulative frequency ratio) is calculated as the division rule evaluation value.

  When the input attribute condition determined by the input attribute condition determining unit 11 is of the type “xj ≦ xj−th” (when the first xj frequency cumulative difference is extracted by the input attribute condition determining unit 11) ), The frequency cumulative lower ratio (= B / A) is calculated as the division rule evaluation value. Here, the frequency cumulative lower ratio (= B / A) can be detected separately from the non-defective first data group DA1 and the defective second data group DA2 according to the input attribute condition “xj ≦ xj−th”. Represents a percentage.

  When the input attribute condition determined by the input attribute condition determining unit 11 is of the type “xj> xj−th” (when the second xj frequency cumulative difference is extracted by the input attribute condition determining unit 11) The frequency cumulative upper ratio (= (100−B) / (100−A)) is calculated as the division rule evaluation value. Here, the cumulative frequency ratio (= (100−B) / (100−A)) is separated from the non-defective first data group DA1 according to the input attribute condition “xj> xj−th”. The ratio at which the second data group DA2 can be detected is shown.

  In other words, the division rule evaluation value (frequency cumulative lower ratio or frequency cumulative upper ratio) is “data included in the second data group DA2 if the input attribute xj satisfies the input attribute condition” for each input attribute condition. The probability of the second association rule “is” is expressed.

  Table 37 shows the division rule evaluation values (frequency cumulative lower ratio or frequency cumulative upper ratio) for each input attribute condition.

[Step 12]
The data dividing unit 15 extracts an input attribute condition that maximizes the value among the division rule evaluation values (frequency cumulative lower ratio or frequency cumulative upper ratio: Table 37) in step 11 above.

  Referring to Table 37, the input attribute condition “x2> 2” has the largest division rule evaluation value among all the input attribute conditions, and division rule evaluation value = frequency cumulative upper ratio = ∞. This indicates that, under the input attribute condition “x2> 2,” the defective second data group DA2 can be detected by being completely separated from the non-defective first data group DA1.

  From another point of view, the input attribute condition “x2> 2” corresponds to the defective second data group DA2 regardless of the values of the other input attributes (x1, x3, x4). From determining the input attribute conditions of other input attributes (x1, x3, x4) (step 8), or calculating the input attribute condition evaluation index (xj frequency cumulative difference) (step 7), It may be a disturbance factor. In such a case, it is desirable to obtain the input attribute conditions of the other input attributes (x1, x3, x4) by excluding data corresponding to “x2> 2” from the analysis data group DA00 ′.

  Therefore, the data dividing unit 15 satisfies the analysis data group DA00 ′ based on the extracted input attribute condition “x2> 2”, the factor data group satisfying “x2> 2”, and “x2> 2”. It is divided into other data groups that do not exist (satisfying “x2 ≦ 2”). Table 38 shows the factor data group, and Table 39 shows the other data group.

  Here, the frequency cumulative lower ratio or the frequency cumulative upper ratio is calculated as the division rule evaluation value, but other evaluation indexes, for example, the improvement degree of the Gini variance index described in Non-Patent Document 1, the information gain (Gain) may be used.

[Step 13]
The analysis data group extraction unit 6 extracts another data group from the data group divided in step 12 as the next analysis data group DA00 ′. Then, the series of processing from step 3 to step 12 is repeated until the end condition determining unit 16 determines that the end condition is satisfied. The end condition determination unit 16 according to the present embodiment determines that the end condition is a case where the number of data in the second data group DA2 of defective products becomes 0 in the above-described step 4 during the iterative process. 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.

  Note that 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 in the above step 4 during the repetitive processing. (2) When the ratio of the number of data in the second data group DA2 to the number of data in the first data group DA1 is equal to or less than a predetermined ratio in the above step 4 during the iterative process, (3) step 12 during the iterative process. The division rule evaluation value of the input attribute condition extracted in (1) may be lower than a predetermined 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.

  Table 40 (corresponding to Table 35 for the first time) shows the calculation result of the defective product separation degree calculation unit 12 in Step 9 for the second time. In the example of Table 40, the input attribute condition of x1, x3, and x4 has a defective product separation degree (second data group separation degree) higher than the defective product content rate (second data group content rate) of the basic data group DA00 after classification. These conditions are extracted (second step 10: Table 41). On the other hand, since the defective product separation degree (second data group separation degree) under the input attribute condition (x2 ≦ 1) of x2 is lower than the defective product content rate of the basic data group DA00 after classification, the input attribute of x2 Condition is not extracted.

  As described above, the problem event (the second data group DA2 of defective products) is obtained by the second processing (processing in which the data satisfying “x2 ≦ 2” in the basic data group DA00 after classification is set as the analysis data group). As input factors, input attribute conditions of “x1> 2”, “x3> 2”, and “x4 ≦ 10” were extracted (Table 41).

  Further, the division rule evaluation values (frequency cumulative lower ratio or frequency cumulative upper ratio) calculated in the second step 11 are shown in Table 42 (corresponding to the first table 37). In this example, the division rule evaluation value is a maximum of 4 in “x1> 2” and “x4 ≦ 10”, but the data division unit 15 selects one of these. This selection criterion should have a certain rule. For example, the input attribute xj with a smaller number j is given priority, and “x1> 2” is selected (step 12 in the second time).

  Of the data group divided by the data dividing unit 15, another data group (a data group satisfying “x1 ≦ 2” in the second analysis data group) is an analysis data group extraction unit as the third analysis data group 6 (Table 43). However, since the analysis data group for the third time in Table 43 did not include defective product data (second data group DA2; y = 2), in step 4 for the third time (up to the second factor extraction). ) Repeat processing has been completed.

  A more detailed factor analysis result can be obtained by the repeated processing as described above than when the repeated processing is not performed.

  Further, when the repeated processing is not performed, even if the accuracy of the input attribute condition evaluation index (xj frequency cumulative difference) is low due to the influence of disturbance, this problem can be solved by performing the repeated processing.

Furthermore, in a certain input attribute, when the factor of the output attribute condition corresponding to the second data group of the defective product is two types of “input attribute is below threshold” and “input attribute exceeds threshold” Both factors can be extracted by repeated processing.
[Step 14]
Table 44 shows an extraction factor list table in which the input attribute conditions (Table 35 and Table 41) extracted for each repetition process of Step 10 are summarized. The extracted factor list table of Table 44 shows all of the plurality of input attribute conditions for the same input attribute by the repetition processing of the factor extracting unit 13 (step 10).

  The factor determination unit 17 selects only a condition having a high priority among a plurality of input attribute conditions (Table 44) for the same input attribute.

  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”.

  In the case of the example in Table 44, finally, the four conditions shown in Table 45 were selected as factors of the problem event (defective product second data group DA2). Table 45 is a list of input attribute conditions determined (selected) by the factor determination unit 17 as the cause of the problem phenomenon, and this table is referred to as a determination factor list table. The determination factor list table is stored in the analysis result data storage unit 20.

  As described above, the factor determination unit 17 uses a plurality of input attributes in the same input attribute extracted for each repetition process of step 10 based on a clear index of defective product separation (second data group separation). Since only the high-priority conditions are selected from among the conditions, the above process is repeated in order to obtain a detailed factor analysis result. The cause of the output attribute condition corresponding to the second data group can be determined by the accuracy.

  If the process of the above two iterations is expressed in the form of a decision tree (similar to the conventional decision tree-2 (FIG. 13)), it is as shown in FIG. Referring to FIG. 8, in this embodiment, for each branch of a decision tree, not only a branch condition but also an input attribute condition that causes a problem event (defective product second data group DA2) is obtained for all input attributes. (Processing by the input attribute condition determination unit 11 in step 8) Among these, only input attribute conditions having a high degree of defective product separation are extracted (processing by the factor extraction unit 13 in step 10). 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 defect factor (factor determination in step 14). Processing by the unit 17).

  In this way, even if the condition is other than the branching condition, all the conditions with a high degree of defective product separation are extracted, so even if there is a competing factor in the branching condition, it is reliably captured without missing that factor. be able to. Further, in the factor extraction for each branch (processing by the factor extracting unit 13) and final factor determination (processing by the factor determining unit 17), the factor is extracted or determined based on a clear index of defective product separation. As a result, no matter how complex the decision tree is, the cause of the problem can be clearly understood. Furthermore, since the degree of defective product separation is used as an evaluation index, it is possible to prioritize a plurality of determined factors (input attribute conditions).

  In this example (data analysis using Table 1 as the basic data group DA), the number of data and the number of input attributes are small, and the influence between the input attributes is small. (Table 45) was the same as the processing result (Table 36) of the factor extraction unit 13 in Step 10 for the first time. However, when the number of data and the number of input attributes are large, the input attribute conditions for each repetition process are often complicated as shown in the extraction factor list table of Table 46. Table 46 is an example (dummy sample of the extraction factor list table) in which four different input attribute conditions are extracted for the input attribute xk1 by the four iterations of step 10. In the case of this example, the condition “xk1 ≦ 2.33” that maximizes the defective product separation degree (second data group separation degree) with respect to the first pattern “input attribute is equal to or less than threshold”. However, in the second pattern that “the input attribute exceeds the threshold”, the condition “xk1> 2.44” that maximizes the defective product separation degree (second data group separation degree) is It is selected by the factor determination unit 17 (Table 47).

  Note that if a factor is to be determined simply, the first pattern and the second pattern are not distinguished from each other, and the input attribute condition that maximizes the defective product separation degree is simply selected from the same input attributes. It may be selected (in the example of Table 46, “xk1> 2.44” is selected).

[Step 15]
The complex factor defect count calculation unit 18 calculates the number of defects due to the complex factor of two conditions among the input attribute conditions in the decision factor list table (Table 45) (Table 48). In Table 48, each input attribute condition is shown in the title row and the title column, and the number of defects (the number of the second data group DA2) due to the composite factor of the two input attribute conditions is shown at the intersection. It is shown.

For example, “x1> 2” row and “x2> 2” column are
This represents the number of data (= 1) satisfying “x1> 2” and “x2> 2” and corresponding to the defective second data group DA2. Hereinafter, the table in Table 48 is referred to as a composite factor table.

[Step 16]
The numerical value-character data conversion unit 19 converts the numerical value of the input attribute threshold value xj-th in the determination factor list table (Table 45) and the composite factor table (Table 48) into character data as necessary. The conversion rule for character data is a rule that is the reverse conversion of the conversion in step 0, and is as follows.
(X1) 1 → A, 2 → B, 3 → C, 4 → D
(X2) 1 → a, 2 → b, 3 → c, 4 → d
(X3) Not converted (x4) Not converted Table 49 shows a factor list table in which the input attribute threshold value xj-th in the determination factor list table in Table 45 is converted into character data.

[Step 17]
The data analysis is completed as described above, and the extraction factor list table (Table 44), the determination factor list tables (Table 45 and Table 49), the composite factor table (Table 48) and various information in the data analysis process are finally The analysis result data is stored in the analysis result data storage unit 20 such as a hard disk. The analysis result data is appropriately sent from the analysis result data storage unit 20 to the output unit 21 such as a display device or a printing device, and is displayed as a table (for example, Table 49), a decision tree (for example, FIG. 8) or a graph. It can be displayed on the device or printed on the printing device.

  As an example, FIG. 9 shows an example in which the decision factor list table (Table 49) is displayed as a factor breakdown Pareto diagram. In FIG. 9, the number of defects (number of second data group DA2) due to each input attribute condition is shown by a bar graph, and the degree of defective product separation (second data group separation degree) is shown by a line graph.

  By referring to the result of FIG. 9, the user has a range of values for each of the input attributes x1 to x4. It is possible to judge immediately by looking at the causes of product characteristic defects. Further, the order (priority order) in which measures should be taken can be determined from the defective product separation degree (second data group separation degree). Furthermore, as a result of the countermeasure against the input attribute condition of FIG. 9, it can be estimated from the number of defects (number of second data groups DA2) how much the number of defects can be reduced.

  In the case of the example in FIG. 9, the first measure is the input attribute x2 (“x2> 2”, that is, “x2 = c or d”), which has the highest defective product separation degree (second data group separation degree). This measure should eliminate two of the four defects (50% of the total defects are resolved).

  As for the content to be secondly countermeasured, a composite factor table (Table 48) is used to combine the first factor (“x2> 2”, ie, “x2 = c or d”) with other factors. It can be judged by examining the degree. FIG. 10 shows the number of defects due to a composite factor with the first factor (“x2> 2”, ie, “x2 = c or d”) in the bar graph (number of defects) of each factor (input attribute condition) in FIG. It is shown with hatching. From FIG. 10, “x1> 2”, that is, “x1 = C or D”) has a high degree of defective product separation (second data group separation), and the first factor (“x2> 2”, that is, Since there are many defects that do not overlap with “x2 = c or d”), there is a high possibility of an independent factor with respect to the first factor, and it can be read that this is an item to be secondly addressed.

  In FIG. 10, it is also possible to extract a composite factor (or a dependent factor) with the first factor (“x2> 2”, ie, “x2 = code”). “X4 ≦ 10” having a large ratio is extracted.

As described above, in the present embodiment, the defective product separation degree (the first defective data separation rate (first defective rate when the data belonging to the input attribute condition in the basic data group DA00 after classification is a population)) is shown. The cause of failure is determined based on a clear index of (2 data group separation). Therefore, a plurality of conditions are extracted in the form of a hierarchical structure as in the conventional decision tree-2 (FIG. 13) (product characteristic defects with y = Y appear throughout the decision tree, and the same input Branches occur under conditions with different attributes many times), and the problem of “it is difficult to determine whether the product characteristics are bad because each input attribute is in which value range” is solved. For example, the condition of product characteristic failure y = Y extracted in the conventional decision tree-2 is
・ "X2 = c or d"
Or
“X2 = a or b” and “x3 ≧ 2.5” and “x1 = C or D”
On the other hand, the product characteristic defect condition (input attribute condition) determined in this embodiment is
・ “X1> 2” (C or D)
"X2>2" (c or d)
・ “X3> 2”
・ “X4 ≦ 10”
It is.

  In other words, which is difficult to understand in the conventional decision tree-2 (FIG. 13), whichever is worse, “input attribute x2,“ x2 = c or d ”or“ x2 = a or b ”. It can be clearly determined that “x2 = c or d” is worse than the question “Is it a condition?”.

  In addition, the condition “x1 = C or D” causes a defect only in the case of a combination with the conditions “x2 = a or b” and “x3 ≧ 2.5”. With respect to the question “whether it becomes a factor”, it is possible to make a clear determination that only the condition “x1 = C or D” causes a defect (the degree of defective product separation is high).

  In the conventional decision tree-2, the input attribute x4 is not extracted as a cause of failure, but in this embodiment, “x4 ≦ 10” is extracted as a cause of failure for the input attribute x4. As shown in FIG. 8, 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 (determining the input attribute condition in step 8). Among these, the input attribute condition having a high degree of defective product separation is extracted (processing by the factor extraction unit 13 in step 10).

  Furthermore, in the conventional decision tree-2, the priorities of the extracted conditions are not prioritized, so it is unclear which condition should be taken and how many defects will be eliminated by that measure. Although it is clear, in the present embodiment, since a clear index of defective product separation (second data group separation) is used, prioritization of extracted factors can be performed. This makes it possible to clarify the conditions that should be preferentially dealt with, and it is possible to estimate from the number of defects how many defects are eliminated by the measure.

  Furthermore, since the factor extraction unit 13 has extracted all the conditions with a high degree of defective product separation (second data group separation degree) even under conditions other than the branching condition in the decision tree, there are competing factors in the branching condition. Even if it exists, it can be caught reliably without missing the factor.

  As described above, according to this embodiment, the first object of the present invention can be achieved.

  Further, according to the present embodiment, the xj frequency cumulative difference with respect to each value of the input attribute xj is used as the input attribute condition evaluation index when the non-defective product and the defective product are separated. 13) Even if the label hierarchical structure (FIG. 12) is not defined in advance, as shown in the input attribute condition column of the determinant list table of Table 49 (or Table 45), The cause of the problem event can be derived. Then, using this, it is possible to obtain the defective product separation degree (second data group separation degree) and the number of defects of each factor (input attribute condition) for the problem event (FIG. 9). That is, the second object of the present invention can be achieved.

  In the above-described embodiment, the decision tree (FIG. 8) is generated by a plurality of branches (repetition). However, if only one branch is sufficient, the process may be terminated in the first step 10.

  In the above description, in step 13, the analysis data group extraction unit 6 extracts only the other data group from the divided data groups as the next analysis data group. However, the factor data group is also extracted as the analysis data group. The series of processing from step 3 to step 12 may be repeated. What is necessary is just to extract at least one of a factor data group and another data group as a new analysis data group.

  In the above description, the data analysis example in which the second data group DA2 is the defective data group and the cause of the defect is extracted has been described. However, the second data group DA2 is the good data group and the condition for obtaining the good product is shown. It is good also as data analysis which extracts.

  As described above, the data analysis apparatus according to the present embodiment classifies the basic data group into the first data group and the second data group according to the value of the output attribute, and assigns a classification flag. Part (classification means) 4 and an analysis data group extraction part (analysis data group extraction means) for extracting an analysis data group to be analyzed from the basic data group classified by the data classification part (classification means) 4 6 and for each of the input attribute conditions that each input attribute of the analysis data group can take, “if the input attribute satisfies the input attribute condition, the data belongs to the second data group in the analysis data group, Input attribute condition evaluation index (Xj frequency cumulative difference) representing the probability of the first association rule that the input attribute does not satisfy the input attribute condition is data belonging to the first data group in the analysis data group Play The first evaluation means (data row separation unit 7, data string extraction unit 8, frequency calculation unit 9, frequency cumulative difference calculation unit 10) and the maximum input attribute condition for each input attribute of the analysis data group An input attribute condition determining unit (input attribute condition determining unit) 11 that determines an input attribute condition having an evaluation index as an input attribute condition satisfying the first correlation rule, and an input attribute condition determining unit (input attribute condition determining unit) 11 For each of the input attribute conditions determined in (2), a second data group separation degree is calculated, which represents a 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. Among the input attribute conditions determined by the defective product separation degree calculation unit (second data group separation degree calculation unit) 12 and the input attribute condition determination unit (input attribute condition determination unit) 11, they are included in the basic data group. First The input attribute condition having the second data group separation degree having a value larger than the second data group content ratio representing the ratio of the number of data in the data group is used as information indicating the cause of the output attribute condition corresponding to the second data group A factor extracting unit (factor extracting means) 13 for extracting is included.

  According to this, the data classification unit 4 classifies the basic data group into the first data group and the second data group, and the analysis data group extraction unit 6 selects the input attribute and the output attribute from the basic data group after the classification. Analytical data group to be analyzed for the causal relationship with. For example, the first data group is a data group having a non-defective product output attribute, and the second data group is a data group having an output attribute representing a problem event such as a defective product.

  The first evaluation means including the data row separation unit 7, the data string extraction unit 8, the frequency calculation unit 9, and the frequency cumulative difference calculation unit 10 is configured so that all input attribute conditions that each input attribute of the analysis data group can take "If the input attribute satisfies the input attribute condition, the data belongs to the second data group in the analysis data group, and if the input attribute does not satisfy the input attribute condition, the first data in the analysis data group. The input attribute condition evaluation index (Xj frequency cumulative difference) representing the probability of the first association rule that “the data belongs to the data group” is calculated, and the input attribute condition determination unit 11 calculates each input attribute of the analysis data group For each, the input attribute condition having the maximum input attribute condition evaluation index is determined as the input attribute condition satisfying the first correlation rule. The defective product separation degree calculation unit 12 calculates a second data group separation degree for each of the input attribute conditions determined by the input attribute condition determination unit 11.

  The second data group separation degree of each input attribute condition represents 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. Of the input attribute conditions determined by the input attribute condition determination unit 11, the second data having a value larger than the second data group content ratio representing the ratio of the number of data of the second data group included in the basic data group An input attribute condition having a group separation degree is extracted as information indicating a factor of an output attribute condition corresponding to the second data group.

  In this way, the data analysis apparatus of this embodiment analyzes the causal relationship between the input attribute and the output attribute for the basic data group, and extracts information indicating the causal relationship.

  In the above causal analysis, the factor extracting unit 13 selects the second data group from the input attribute conditions determined by the input attribute condition determining unit 11 based on a clear index of the second data group separation degree. The factor of the corresponding output attribute condition, that is, the input attribute condition that makes the data the second data group is extracted. Therefore, no matter how complex the decision tree is, it is possible to clearly extract the factor of the output attribute condition corresponding to the second data group. For example, if the output attribute of the second data group represents a problem phenomenon such as a defective product, the second data group separation degree is a defective product separation degree indicating the accuracy of defective product extraction, and how complex the decision tree is. It is possible to clearly understand the cause of the problem phenomenon.

  Further, since the second data group separation degree such as the defective part separation degree is used as the evaluation index, it is possible to prioritize a plurality of factors (input attribute conditions) extracted by the factor extracting unit 13. .

  Furthermore, since the factor extraction unit 13 extracts all conditions having a high second data group separation degree such as a defective product separation degree, even if the condition is other than the branching condition in the decision tree, there are competing factors in the branching condition. Even if it exists, it can be caught reliably without missing the factor.

As described above, a factor of a predetermined output attribute can be extracted based on a clear index.
In addition, the data analysis apparatus according to the present embodiment, for each of the input attribute conditions determined by the input attribute condition determination unit 11, “if the input attribute satisfies the input attribute condition, the second data in the analysis data group A frequency cumulative ratio calculation unit (second evaluation means) that calculates a division rule evaluation value (frequency cumulative lower ratio or frequency cumulative upper ratio) representing the probability of the second correlation rule that the data is included in the group ) 14 and the factor data group satisfying the input attribute condition based on the input attribute condition having the largest division rule evaluation value among the input attribute conditions determined by the input attribute condition determining unit 11 And a data dividing unit (dividing unit) 15 that divides the data into other data groups that do not satisfy the input attribute condition. The analysis data group extracting unit 6 includes the factor data group divided by the data dividing unit 15 and At least one of the data groups is extracted as a new analysis data group, processing by the analysis data group extraction unit 6, first evaluation means (data row separation unit 7, data string extraction unit 8, frequency calculation unit 9, frequency accumulation) Processing by the difference calculating unit 10), processing by the input attribute condition determining unit 11, processing by the defective product separation degree calculating unit 12, processing by the factor extracting unit 13, processing by the frequency cumulative ratio calculating unit 14, and by the data dividing unit 15 A series of processes consisting of processes is repeatedly executed.

  According to this, for each of the input attribute conditions determined by the input attribute condition determination unit 11 by the frequency cumulative ratio calculation unit 14, “if the input attribute satisfies the input attribute condition, the second data group in the analysis data group The division rule evaluation value representing the probability of the second correlation rule that is “data included in the data” is calculated, and the data division unit 15 calculates the analysis data group based on the input attribute condition having the largest division rule evaluation value. Are divided into a factor data group and another data group, and the analysis data group extraction unit 6 extracts at least one of the factor data group and the other data group as a new analysis data group. Then, the above series of processing is repeatedly executed.

  More detailed factor analysis results can be obtained by such repeated processing.

Further, when the repeated processing is not performed, even if the accuracy of the input attribute condition evaluation index is low due to the influence of disturbance, this problem can be solved by performing the repeated processing.
Furthermore, in a certain input attribute, the output attribute condition factor corresponding to the second data group is repeated even when the two types are “input attribute is below threshold” and “input attribute exceeds threshold”. Both of these factors can be extracted by this process.
In addition, the data analysis apparatus according to the present embodiment is such that only a condition having a high priority is selected among a plurality of input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting unit 13. Is provided with a factor determining unit (factor determining means) 17 for selecting.

  Therefore, even if the above iterative process is performed to obtain a detailed factor analysis result, the factor of the output attribute condition corresponding to the second data group can be determined in a very simple form without complication. .

  Further, in the data analysis apparatus according to the present embodiment, the factor determination unit 17 determines the defective product separation degree among the plurality of input attribute conditions for the same input attribute extracted by the repeated processing of the factor extraction unit 13. The input attribute condition that maximizes (second data group separation degree) is selected as a condition with high priority.

  According to this, for a plurality of input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting unit 13, a condition having a high priority is set based on a clear index of the second data group separation degree. Since the selection is made, it is possible to determine the factor of the output attribute condition corresponding to the second data group with high accuracy while being in a very simple form.

  Further, the factor determination unit 17 outputs the output attribute corresponding to the second data group from the input attribute conditions extracted by the iterative process of the factor extraction unit 13 based on the clear index of the second data group separation degree. The factor of the condition, that is, the input attribute condition for which the data becomes the second data group is determined. Therefore, no matter how complicated the decision tree is, it is possible to clearly determine the factor of the output attribute condition corresponding to the second data group. For example, if the output attribute of the second data group represents a problem such as a defective product as described above, the second data group separation degree is a defective product separation degree indicating the accuracy of defective product extraction, and how complicated it is. Even if it becomes a decision tree, it is possible to clearly grasp the cause of the problem phenomenon.

  Further, since the second data group separation degree such as the defective part separation degree is used as an evaluation index, it is possible to prioritize a plurality of factors (input attribute conditions) determined by the factor determining unit 17. .

  Furthermore, in the process of the above iterative process, the factor extraction unit 13 extracts all conditions having a high second data group separation degree, such as a defective product separation degree, even under conditions other than the branch condition in the decision tree. . The factor determination unit 17 outputs the output attribute condition corresponding to the second data group based on a clear index of the second data group separation degree from the input attribute conditions extracted by the repetition processing of the factor extraction unit 13. Factor, that is, the input attribute condition for which the data becomes the second data group is determined, so that even if there is a competing factor in the branch condition, the output corresponding to the second data group can be surely performed without missing the factor. Factors for attribute conditions can be determined.

  Further, in the data analysis apparatus according to the present embodiment, a plurality of input attribute conditions in the same input attribute extracted by the repeated processing of the factor extraction unit 13 are “first input attribute is equal to or less than threshold”. And the second pattern that the input attribute exceeds the threshold value, the factor determination unit 17 uses the second data group separation degree such as the defective product separation degree in the first pattern. One input attribute condition that maximizes the second data group and one input attribute condition that maximizes the second data group separation degree in the second pattern are selected as conditions with high priority.

  According to this, the plurality of input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting unit 13 are “input attribute is below threshold” and “input attribute exceeds threshold” Even in the case of the two patterns, it is possible to determine the factor of the output attribute condition corresponding to the second data group with high accuracy while having a very simple form.

  The analysis data group extraction unit 6 of the data analysis apparatus according to the present embodiment extracts only other data groups from the data group divided by the data division unit 15 as new analysis data groups.

  According to this, since only the other data group among the data groups divided by the data dividing unit 15 is subjected to the above-described repetitive processing as a new analysis data group, the output attribute condition corresponding to the second data group A simple and sufficient factor analysis result is obtained for the factor analysis.

  In addition, since the other data group is processed as a new analysis data group, the analysis can be performed by excluding the influence of the input attribute condition that satisfies the second correlation rule in the previous iteration process, Thereby, a new factor of the output attribute condition corresponding to the second data group can be extracted with high accuracy.

  Note that the data analysis apparatus according to the present embodiment includes an end condition determination unit (end condition determination unit) 16 that determines whether or not the end condition is satisfied, and the end condition determination unit 16 satisfies the end condition. If it is determined, execution of the series of processes is terminated.

  According to this, since the above repetitive processing is performed while performing the determination by the end condition determination unit 16, it is possible to omit the repetitive processing more than necessary to obtain a predetermined factor analysis result.

  Then, the end condition determination unit 16 determines whether the number of data in the second data group in the analysis data group extracted by the analysis data group extraction unit 6 is 0 as an end condition.

  According to this, since the process is repeatedly performed until the number of data in the second data group in the analysis data group becomes zero, a detailed factor analysis result can be obtained.

  Further, in the data analysis apparatus according to the present embodiment, all the input attributes in the analysis data group are numerical attributes, and the first evaluation unit performs the first analysis data group for all the numerical values of each input attribute. In the data group, the ratio of the number of data whose input attribute is less than or equal to the numerical value is calculated as the first frequency, and in the second data group of the analysis data group, the ratio of the number of data whose input attribute is equal to or less than the value Is calculated as a second frequency, and a frequency cumulative difference calculation unit 10 that calculates the difference between the first frequency and the second frequency for all the numerical values of each input attribute is included.

  According to this, the difference (Xj frequency cumulative difference) between the first frequency and the second frequency calculated by the frequency cumulative difference calculation unit 10 for each numerical value of the input attribute is expressed as “If the input attribute is less than or equal to the numerical value. Data belonging to the second data group. If the input attribute exceeds the numerical value, the data belongs to the first data group. "Or" If the input attribute exceeds the numerical value, the second data group. It can be used as an input attribute condition evaluation index representing the probability of the first association rule that the data belongs to the data and belongs to the first data group if the input attribute is equal to or less than the numerical value.

  Thus, since the difference between the first frequency and the second frequency is used as an input attribute condition evaluation index when the first data group and the second data group are separated, the label hierarchical structure is defined in advance. It is possible to determine an input attribute condition that satisfies the first correlation rule without any trouble and with simple processing. Accordingly, it is possible to provide a data analysis apparatus that can derive a causal relationship between an output attribute and an input attribute in a simple manner without defining a label hierarchical structure in advance.

  The data analysis apparatus according to the present embodiment includes a character-numeric data conversion unit (numerical conversion means) 2 that performs numerical conversion processing on non-numeric data in the basic data group.

  According to this, since numerical conversion processing is performed on non-numeric data in the basic data group, even if non-numeric data is included in the basic data group, the frequency calculation unit 9 and the frequency cumulative difference Processing by the first evaluation unit including the calculation unit 10 can be performed.

  In the data analysis apparatus according to the present embodiment, the frequency cumulative ratio calculation unit 14 inputs the input attribute conditions determined by the input attribute condition determination unit 11 in the first data group of the analysis data group. The ratio of the ratio of the number of data satisfying the input attribute condition in the second data group of the analysis data group to the ratio of the number of data satisfying the attribute condition is calculated as the division rule evaluation value.

  According to this, for each of the input attribute conditions determined by the input attribute condition determination unit 11, the ratio that can be detected from the first data group by the input attribute condition and the second data group can be calculated as the division rule evaluation value can do.

  The data analysis apparatus according to the present embodiment includes a classification condition setting unit 3 that sets classification conditions, and the data classification unit 4 classifies the basic data group based on a comparison between the value of the output attribute and the classification conditions. It is supposed to be.

  According to this, the basic data group can be classified into the first data group and the second data group based on a predetermined classification condition set as appropriate. That is, an output attribute condition corresponding to the second data group can be set as appropriate, and the factor can be extracted or determined.

  In the data analysis apparatus according to the present embodiment, the basic data group includes a plurality of output attributes, and the classification condition setting unit 3 sets classification conditions for each of the plurality of output attributes, and the data classification unit For each output attribute, the value of the output attribute is compared with the corresponding classification condition, and the basic data group is classified by the logical sum or logical product of the comparison results in each output attribute. .

  According to this, even when the second data group is defined by a plurality of output attribute conditions, the basic data group can be classified into the first data group and the second data group. Thereby, the factor of the output attribute condition corresponding to the second data group defined by a plurality of output attribute conditions can be extracted or determined.

  The data analysis method described above can be realized by the computer executing a data analysis program including processes corresponding to S0 to S17 (steps 0 to 17) in FIG. Therefore, in the data analysis apparatus of FIG. 1, the data analysis program includes a computer, a character-numeric data conversion unit 2, a classification condition setting unit 3, a data classification unit 4, an analysis data group extraction unit 6, a data row separation unit 7, Data string extraction unit 8, frequency calculation unit 9, frequency cumulative difference calculation unit 10, input attribute condition determination unit 11, defective product separation degree calculation unit 12, factor extraction unit 13, frequency cumulative ratio calculation unit 14, data division unit 15, This can be realized by functioning as an end condition determination unit 16, a factor determination unit 17, a complex factor defect number calculation unit 18, and a numerical value-character data conversion unit 19.

  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. In addition, as the removable media, optical recording media such as CD-ROM and DVD; magneto-optical recording media such as MO; magnetic recording media such as floppy (registered trademark) disk, cassette tape, and removable hard disk; memory cards and the like And a medium having a built-in rewritable nonvolatile memory such as a medium having a built-in 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. In addition, the program is not limited to be sold in a state where it is stored in a computer-readable recording medium, and 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.

  The present invention can be applied to analysis of the causal relationship between an output attribute (object attribute) to be analyzed and an input attribute (description attribute) that is an attribute affecting the output attribute.

It is a block diagram which shows the structure of the data analyzer which concerns on one Embodiment of this invention. It is a flowchart which shows the data analysis method which concerns on one Embodiment of this invention. The graph of an example of the output of the frequency accumulation difference calculating part 10 (step 7) in the data analysis apparatus which concerns on one Embodiment of this invention is represented with the value of the input attribute x1, and 1-x1 frequency accumulation% ( A) shows the relationship between 2-x1 frequency cumulative% (B) of defective products, first x1 frequency cumulative difference (BA), and second x1 frequency cumulative difference (AB). The graph of an example of the output of the frequency accumulation difference calculating part 10 (step 7) in the data analyzer which concerns on one Embodiment of this invention is represented with the value of the input attribute x2, and 1-x2 frequency accumulation% (non-defective product). A) shows the relationship between 2-x2 frequency cumulative percentage (B) of defective products, first x2 frequency cumulative difference (BA), and second x2 frequency cumulative difference (AB). FIG. 6 is a graph showing an example of the output of the frequency cumulative difference calculation unit 10 (step 7) in the data analysis apparatus according to the embodiment of the present invention, and the value of the input attribute x3 and the non-defective 1-x3 frequency cumulative% ( A) shows the relationship between 2-x3 frequency cumulative% (B) of defective products, first x3 frequency cumulative difference (BA), and second x3 frequency cumulative difference (AB). FIG. 6 is a graph showing an example of the output of the frequency cumulative difference calculation unit 10 (step 7) in the data analysis apparatus according to the embodiment of the present invention, and the value of the input attribute x4 and the non-defective 1-x4 frequency cumulative% ( A) shows the relationship between 2-x4 frequency cumulative% (B) of defective products, first x4 frequency cumulative difference (BA), and second x4 frequency cumulative difference (AB). It is the figure which expressed an example (Table 35) of the data output by the inferior goods isolation | separation degree calculating part 12 (step 9) in the data analyzer which concerns on one Embodiment of this invention with the Venn diagram. It is the figure which expressed the process of the factor extraction (step 10) and the factor determination (step 14) of the data analysis method which concerns on one Embodiment of this invention in the form of a decision tree. With respect to an example of the determination factor list table (Table 49) output by the factor determination unit 17 (step 14) in the data analysis apparatus according to the embodiment of the present invention, the number of defects for each input attribute condition is represented by a bar graph and defective products are separated. It is the figure which expressed degree (2nd data group separation degree) with the line graph. Using an example (Table 48) of data output from the composite factor defect number calculation unit 18 (step 15) in the data analysis apparatus according to an embodiment of the present invention, the number of defects for each input attribute condition is represented by a bar graph. The degree of separation (second data group separation degree) is expressed in a line graph, and hatching is applied to the number of defects due to a combined factor with the first factor (“x2> 2”, ie, “x2 = c or d”). It is attached. It is a figure showing the conventional decision tree-1. It is a figure showing the label hierarchical structure of the conventional decision tree-2, (a) shows x1 attribute, (b) shows x2 attribute, (c) shows x3 attribute, (d) shows x4 attribute. It is a figure showing the conventional decision tree-2.

Explanation of symbols

1 Basic data group storage unit 2 Character-numeric data conversion unit (numeric conversion means)
3 Classification condition setting part (Classification condition setting means)
4 Data classification part (classification means)
5 Basic data group storage after classification 6 Analysis data group extraction unit (analysis data group extraction means)
7 data row separation unit 8 data string extraction unit 9 frequency calculation unit (first evaluation means, frequency calculation means)
10 Frequency cumulative difference calculation unit (first evaluation means, difference calculation means)
11 Input attribute condition determining unit (input attribute condition determining means)
12 Defective product separation degree calculation unit (second data group separation degree calculation means)
13 Factor extraction unit (factor extraction means)
14 Frequency cumulative ratio calculation unit (second evaluation means)
15 Data division unit (division means)
16 End condition determination unit (end condition determination means)
17 Factor determination unit (factor determination means)
18 Complex factor defect count calculation unit 19 Numerical value-character data conversion unit 20 Analysis result data storage unit 21 Output unit

Claims (11)

Analyzing the causal relationship between the input attribute and the output attribute with respect to a basic data group that is a set of data composed of a plurality of numeric type input attributes and output attributes, information indicating the causal relationship A data analysis device for extraction,
The basic data group is classified into a first data group and a second data group based on a comparison between the value of the output attribute and a predetermined classification condition, and a classification flag corresponding to the classification result is set as the basic data. A classification means to be given to the group ;
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;
For every numerical value that can be taken by the input attribute of each of the analysis data groups, all the data belonging to the first data group of the number of data belonging to the first data group among the data having numerical values equal to or smaller than the numerical value. Data that belongs to the second data group among data having a numerical value equal to or lower than the numerical value for every numerical value that can be taken by each of the input attributes, and performing an operation for obtaining a first frequency that is a ratio to the number of Frequency calculating means for calculating a second frequency, which is a ratio of the number of data to the number of all data belonging to the second data group;
Difference calculation means for calculating an input attribute condition evaluation index representing a difference value between the first frequency and the second frequency for each of all the numerical values taken by the input attribute;
Based on the numerical value that maximizes the difference value among the numerical values taken by one input attribute, for each of the input attributes, “if the input attribute satisfies the input attribute condition, a data belonging to the second data group, said input attributes to meet the input attribute conditions, satisfy the first is a data belonging to a data group "referred to as a first correlation rule in the analysis data set an input attribute condition determining means that determine the input attribute conditions,
For each of the input attribute conditions determined by the input attribute condition determining means, a 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, A second data group separation degree computing means for computing two data group separation degrees;
Among the input attribute conditions determined by the input attribute condition determining means, a value larger than the second data group content rate representing the ratio of the number of data of the second data group included in the basic data group, A data analysis apparatus comprising: factor extracting means for extracting an input attribute condition having the second data group separation degree as information indicating a factor of an output attribute condition corresponding to the second data group. For each of the input attribute conditions determined by the input attribute condition determining means, if the second frequency in the numerical value of the input attribute applied to the determination of the input attribute condition is greater than the first frequency, the first When the first ratio, which is the ratio of the second frequency to the frequency, is calculated as the evaluation value and the first frequency is greater than the second frequency, (100% −first frequency) is set to (100 % -Second frequency), an evaluation value calculating means for calculating a second ratio as an evaluation value;
Among the input attribute condition determined by the input attribute condition determining means, based on the input attribute conditions with maximum the Review value, the analysis data groups, and input attributes satisfy factor data group, the A dividing unit that divides the data into other data groups that do not satisfy the input attribute condition;
The analysis data group extraction means extracts at least one of the factor data group and the other data group divided by the division means as a new analysis data group, and the processing by the analysis data group extraction means, the frequency Processing by the computing means, processing by the difference computing means , processing by the input attribute condition determining means, processing by the second data group separation degree computing means, processing by the factor extracting means, processing by the evaluation value computing means , and The data analysis apparatus according to claim 1, wherein a series of processing including processing by the dividing unit is repeatedly executed. For the plurality of input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting means, the apparatus further comprises factor determining means for selecting only a condition having a higher priority among them. ,
The factor determining means includes
Among the plurality of input attribute conditions for the same input attribute extracted by the repeated processing of the factor extracting means, an input attribute condition that maximizes the second data group separation degree is selected as the high priority. The data analysis apparatus according to claim 2, wherein the data analysis apparatus is selected as a condition. For the plurality of input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting means, the apparatus further comprises factor determining means for selecting only a condition having a higher priority among them. ,
A plurality of the input attribute conditions in the same input attribute extracted by the repeated processing of the factor extracting means include a first pattern that “the input attribute is equal to or less than a threshold value” and “the input attribute is a threshold value” In the case of being divided into the second pattern of “exceeding”,
The factor determining means includes
One input attribute condition that maximizes the second data group separation degree in the first pattern, and one input attribute condition that maximizes the second data group separation degree in the second pattern The data analysis apparatus according to claim 2, wherein one is selected as a condition having a high priority.   The data according to claim 2, wherein the analysis data group extraction unit extracts only the other data group from the data group divided by the division unit as a new analysis data group. Analysis equipment. It further includes an end condition determining means for determining whether or not the end condition is satisfied,
When the end condition determining means determines that the end condition is satisfied, the execution of the series of processes ends.
3. The end condition determination unit determines whether the number of data in the second data group in the analysis data group extracted by the analysis data group extraction unit is 0 as the end condition. The data analysis device described in 1. It further includes a classification condition setting means for setting a classification condition,
3. The data analysis apparatus according to claim 1, wherein the classification unit classifies the basic data group based on a comparison between the value of the output attribute and the classification condition. A data analysis method for analyzing the causal relationship and extracting information indicating the causal relationship with respect to the basic data group using the data analysis device according to claim 1,
The classification means classifies the basic data group into the first data group and the second data group based on a comparison between the output attribute value and a predetermined classification condition, and corresponds to the classification result. A classification step of assigning the classification flag to the basic data group ;
An analysis data group extraction step of extracting the analysis data group from the basic data group after classification by the classification means by the analysis data group extraction means;
First data of the number of data belonging to the first data group among the data having a numerical value equal to or lower than the numerical value for every numerical value that can be taken by each input attribute of the analysis data group by the frequency calculation means. An operation for obtaining a first frequency which is a ratio to the number of all data belonging to the group, and for each of all the numerical values that can be taken by each of the input attributes, A frequency calculating step for calculating a second frequency that is a ratio of the number of data belonging to the two data groups to the number of all data belonging to the second data group;
A difference calculating step of calculating an input attribute condition evaluation index representing a difference value between the first frequency and the second frequency for each of all the numerical values taken by the input attribute by the difference calculating means;
According to the input attribute condition determining means , for each of the input attributes based on the numerical value that maximizes the difference value among the numerical values taken by one input attribute, If the condition is satisfied, the data belongs to the second data group in the analysis data group, and if the input attribute does not satisfy the input attribute condition, the data belongs to the first data group in the analysis data group. an input attribute condition determination step that determine the input attribute conditions satisfying the first correlation rule,
A second data group separation degree calculating step of calculating the second data group separation degree for each of the input attribute conditions determined by the input attribute condition determining means by the second data group separation degree calculating means;
Among the input attribute conditions determined by the input attribute condition determining means by the factor extracting means, the input attribute condition having the second data group separation degree having a value larger than the second data group content rate. And a factor extracting step of extracting the information as information indicating the factor of the output attribute condition corresponding to the second data group. The input attribute is a manufacturing process condition and / or an in-line inspection result in a product manufacturing process,
The output attribute is a product quality determination result,
The data analysis method according to claim 8, wherein the second data group is a data group in which the quality determination result is defective. Computer
A basic data group, which is a set of data composed of a plurality of input attributes and output attributes, is divided into a first data group and a second data group by comparing the value of the output attribute with a predetermined classification condition. Classification means for classifying and assigning a classification flag corresponding to the classification result to the basic data group ;
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;
For every numerical value that can be taken by the input attribute of each of the analysis data groups, all the data belonging to the first data group of the number of data belonging to the first data group among the data having numerical values equal to or smaller than the numerical value. Data that belongs to the second data group among data having a numerical value equal to or lower than the numerical value for every numerical value that can be taken by each of the input attributes, and performing an operation for obtaining a first frequency that is a ratio to the number of Frequency calculating means for calculating a second frequency, which is a ratio of the number of data to the number of all data belonging to the second data group;
Difference calculation means for calculating an input attribute condition evaluation index representing a difference value between the first frequency and the second frequency for each of all the numerical values taken by the input attribute;
Based on the numerical value that maximizes the difference value among the numerical values taken by one input attribute, for each of the input attributes, “if the input attribute satisfies the input attribute condition, a data belonging to the second data group, said input attributes to meet the input attribute conditions, satisfy the first is a data belonging to a data group "referred to as a first correlation rule in the analysis data set an input attribute condition determining means that determine the input attribute conditions,
For each of the input attribute conditions determined by the input attribute condition determining means, a 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, A second data group separation degree computing means for computing two data group separation degrees;
Among the input attribute conditions determined by the input attribute condition determining means, a value larger than the second data group content rate representing the ratio of the number of data of the second data group included in the basic data group, A data analysis program for functioning as a factor extracting unit that extracts an input attribute condition having the second data group separation degree as information indicating a factor of an output attribute condition corresponding to the second data group. The computer-readable recording medium which recorded the data analysis program of Claim 10.

Images