JP2002324206A - Data analyzing method and its device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data analyzing method capable of clarifying the level of confounding among a plurality of explanation variables. SOLUTION: This data analyzing method comprises a step for preparing the data results of explanation variables and purpose variables, a step for calculating the level of confounding and/or the level of independence among those explanation variables based on the data results, and a step for executing data mining by using the level of confounding and/or the level of independence.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data analysis for grasping a relation between data handled in a wide range of industries, extracting a significant result for providing an industrially superior result, and further analyzing the result. The present invention relates to a method and an apparatus for evaluating the accuracy and the like of a device.

For example, using the history of used equipment, test results, design information, various measurement data, and the like obtained in a semiconductor manufacturing process, the fluctuation state of the yield is grasped, and conditions advantageous for improving the yield are extracted. Data analysis,
And a method and apparatus for evaluating the accuracy and the like of the result.

[0003] In particular, a method and a method for obtaining a more efficient and reliable analysis result by coping with a case where a plurality of explanatory variables are entangled with each other (become independant) and it becomes difficult to extract a significant difference. Related to the device.

[0004]

2. Description of the Related Art The yield analysis of semiconductor data will be described as an example. In particular, when trying to obtain reference data for deciding measures for improving quality and productivity based on the analysis results, as in the case of process data analysis, the accuracy and reliability are important. (Japanese Patent Application No. 2000-41896). In order to find the cause of the decrease in yield as soon as possible and take countermeasures, it is necessary to find out factors that are affecting the yield and other factors that are affecting the factors from the device history, test results, design information, various measurement data, etc. Data analysis is performed. In data analysis, a target variable such as a yield value is an objective variable, and a device history, a test result, design information, various measurement data, and the like, which are factors of the objective variable, are called explanatory variables. At this time, various statistical methods are applied, and by applying data mining as one of them, valuable information and regularity that are difficult to distinguish from a large amount of various data can be extracted.

However, when data mining is actually applied, it sometimes fails. In applications such as finance and distribution, there are millions of huge data records, and the number of explanatory variables is at most tens, so highly accurate analysis results were obtained. However, in the case of semiconductor process data analysis, although the number of data items is small and at most about 200 lots of the same type, the number of explanatory variables reaches several hundreds (equipment history, in-process inspection values, etc.) In some cases, independent variables are no longer independent and simple data mining may not provide reliable results.

[0006]

A description will be given by taking a yield analysis of semiconductor data as an example. Number of data (example: lot number)
In a process data analysis with many explanatory variables (eg, LSI manufacturing process data) compared to the above, a plurality of explanatory variables are entangled with each other (they are no longer independent), and the problem due to statistical significance cannot be sufficiently narrowed down. There are many. Even when data mining (regression tree analysis, etc.) is applied, if this problem exists, it is necessary to check the accuracy of the analysis result and the reliable range with considerable effort.

FIG. 1 shows a relationship between a lot flow and an abnormal manufacturing apparatus. A white circle “○” indicates the normal device 101, and a black circle “●” indicates the abnormal device 102. Arrows indicate the flow of lots. The analysis of the difference between the devices in the LSI manufacturing data extracts, from the used device data for each process of each lot, which manufacturing device is used in which manufacturing process most affects the yield.

FIG. 2 shows a yield distribution (box whisker diagram) for each apparatus in a certain process according to the prior art. The yield value of the lot for each device used in each manufacturing process is displayed in a box-and-whisker diagram, and the process is confirmed, and the process with the most significant difference and the device are identified.

However, this method requires a large number of steps today when the number of steps is several hundred, and it is difficult to make a judgment when a difference is not clearly seen or when conditions are complicatedly entangled. In order to deal with these, a data mining method based on regression tree analysis is effective, and the target variable is divided into a group of used devices having a high value and a group of used devices having a low value of the objective variable.
As shown in FIG. 3, when a device used for each lot is fixed and a lot is flowed, the abnormal device 102 indicated by a black circle “●” may not be uniquely identified. That is, when the independence between explanatory variables is low, a case where the significant difference due to the two divisions of the set is large is not always “true significant difference”.

The above is the confounding of the used equipment in each process of semiconductor manufacturing. The same applies to the confounding of a set divided into two as a result of regression tree analysis. That is,
The same can be said for a set including a device group in which a high yield occurs and a device group in which a low yield occurs in each process. The confounding of the two divided sets is the same when the explanatory variable is a continuous value.

An object of the present invention is to provide a data analysis method and apparatus which can clarify the degree of confounding between a plurality of explanatory variables.

[0012]

According to one aspect of the present invention, a step of preparing data results of an explanatory variable and an objective variable, and a degree of confounding and / or independence between a plurality of explanatory variables based on the data results. A data analysis method includes a step of calculating a degree and a step of performing data mining using the degree of confounding and / or independence.

By calculating the degree of confounding and / or the degree of independence between a plurality of explanatory variables, the degree of confounding of explanatory variables can be clearly grasped. As a result, the degree of confounding of the explanatory variables can be quantitatively evaluated based on the result of dividing the set of the regression tree analysis into two. It is possible to clarify which explanatory variables need attention.

[0014]

FIG. 17 shows a data analyzer incorporating data mining according to an embodiment of the present invention.
The data mining unit 1703 controls the original data group 1
Based on individual original data extracted from each database 1702 in the database 701, a process of extracting features and regularities hidden in the data is performed, and a rule file 1704 is created. The analysis tool group 1705 has a statistical analysis component 1706, a chart creation component 1707, and the like, and a database 17 based on a rule file 1704.
The original data extracted from the original data 02 is analyzed. The analysis result is fed back to the analysis tool group 1705 and the data mining unit 1703. The data mining unit 1703 performs data mining based on the analysis result of the analysis tool group 1703 and the original data group 1701. The analysis tool group 1705 performs an analysis based on the rule file 1704, individual original data extracted from the database 1702, and its own analysis result. The decision making unit 1708 makes a decision based on the analysis result of the analysis tool group 1705.

When data mining is applied in yield data analysis, a measure for improving the yield is determined based on the data mining result, it is determined whether or not the measure should be implemented, and the effect of the measure is predicted. Or will be. For that purpose, quantitative evaluation and accuracy of data mining results are required.

Of the discriminant tree analysis which is one of the data mining techniques, the regression tree analysis is particularly effective. One of the advantages of regression tree analysis is that results are output as easy-to-understand rules, which can be expressed in a general language or a database language such as SQL language. Therefore, the reliability and accuracy of these results are used effectively to make effective decisions based on the results, and actions (that is, measures, etc.)
Can be caused to occur.

The regression tree analysis will be briefly described. Regression tree analysis targets a set of records consisting of explanatory variables indicating multiple attributes and objective variables affected by the variables,
The attribute and the attribute value that most influence the objective variable are determined. The data mining unit (regression tree analysis engine) outputs rules indicating data characteristics and regularity.

First, regression tree analysis will be described. The processing of the regression tree analysis is realized by repeatedly dividing the set into two based on the parameter value (attribute value) of each explanatory variable (attribute). At the time of the set division, when the sum of squares of the objective variable (yield) before division is S0 and the sum of squares of the yields of the two sets after division is S1 and S2, ΔS expressed by equation (1) becomes maximum. In this way, the explanatory variable of the record to be divided and its parameter value are obtained.

ΔS = S0− (S1 + S2) (1)

The explanatory variables and their parameter values obtained here correspond to the branch points in the regression tree. Thereafter, the same processing is repeated for the divided sets, and the influence of the explanatory variables on the objective variables is examined. The above is a generally well-known method of regression tree analysis. In order to grasp the clarity of set division in more detail, the following parameters (a) to ( d) is also used as a quantitative evaluation of the data mining results.

(A) S ratio This is a reduction ratio of the sum of squares by set division, and is a parameter indicating how much the sum of squares has been reduced by set division.
The smaller this value is, the greater the effect of the set division is, and since the set division is clearly performed, the significant difference is large.

S ratio = ((S1 + S2) / 2) / S0 (2)

(B) t The set is divided into two by the regression tree analysis engine, and is a value for testing the difference between the average (/ X1, / X2) of the two divided sets. Here, “/” indicates an overline. The statistical t-test is a criterion indicating a significant difference between the average values of the objective variables in the divided sets. If the degree of freedom, that is, the number of data is the same, the set is clearly divided as t increases, and the significant difference is large.

At this time, if there is no significant difference in the variance of the divided sets, t is obtained by equation (3), and if there is a significant difference in the variance of the divided sets, t is calculated by equation (4). Ask. Here, N1 and N2 are the divided sets 1 respectively.
And the number of elements in set 2. / X1 and / X2 are the averages of each set after division. S1 and S2 are the sum of squares of the objective variables of each set obtained by dividing a set consisting of a plurality of original data into two.

[0025]

(Equation 1)

[0026]

(Equation 2)

(C) Difference in average value of objective variables of divided sets The larger this value is, the larger the significant difference is.

(D) Number of Data of Each Set Divided The smaller the difference between them is, the smaller the effect of abnormal values (noise) is.

FIG. 4 shows a format of an example of data to be input to the regression tree analysis. Each record has a wafer number unit, and each record has a device 411 used in each manufacturing process, electrical characteristic data 412, and a wafer yield 413. The explanatory variables 401 are used devices 411 and electrical characteristic data 412.
And so on. The target variable 402 is the yield 413. For example, it is assumed that the use device 411 and the electrical characteristic data 412 have an effect on the yield. FIGS. 5 and 6 show a regression tree diagram and a statistical value list for evaluation, which are the results of regression tree analysis based on this data.

FIG. 5 is a regression tree diagram which is a result of regression tree analysis. The root node n0 is divided into two nodes n1 and n2. Node n1 is divided into two nodes n3 and n4. Node n2 is divided into two nodes n5 and n6. Node n6 is divided into two nodes n7 and n8.

FIG. 6 shows the statistical values for evaluating the explanatory variables at the time of the first division. For example, the average value A of the objective variables of the entire set
ve is 75, standard deviation s is 12, and the number of data N is 1000. The lists 601 to 604 respectively include, from the left, the order of the significant difference, the S ratio, the t value, the difference between the average values of the objective variables of the divided sets, the number of data of each divided set, and the attribute name of the divided set ( An explanatory variable), an attribute value (parameter value) of the two divided sets, and a magnitude relationship between the objective variable thereof. The lists 601 to 604 are candidates for grouping based on the value of ΔS shown in the expression (1) of the attribute value (explanatory variable) to be divided, and are arranged in descending order of the significant difference (ΔS). FIG. 5 is a diagram showing a node n based on the first candidate 601.
0 is divided into nodes n1 and n2.

When the set n0 of all the wafers in FIG. 5 is divided into two sets n1 and n2 based on the evaluation value of ΔS in the equation (1), the process A has the most influence on the yield in step A.
Whether to use AM1 or AM2, the latter has better yield. Hereinafter, by repeating the same set division for the divided sets, this regression tree diagram is obtained. For a wafer group using AM2 in step A and CM2 in step C, the state where the electrical characteristic data RSP is 90 or less is most effective (high yield).

FIG. 7 is equivalent to FIG. 5, and shows the correlation between the yield of the divided wafer set, the equipment used in the specific process, and the electrical characteristic data. The explanatory variables appearing in the upper hierarchy in the regression tree diagram of FIG. 5 have a greater influence on the objective variables. Although the average yield of all wafers is 74.8%, regression tree analysis automatically shows that these features and regularities are present in several sets in relation to the equipment used and electrical characteristic data. Extraction and provide clues for yield analysis.

In the regression tree diagram of FIG. 5, the upper two hierarchies are all due to the difference in the used equipment. Therefore, in the analysis using all the wafers, it is the difference in the used equipment that has a large influence on the yield even if the combined condition is included. . The electrical characteristics data seem to be less effective. However, it can be seen from FIGS. 5 and 7 that the RSP is most effective for the yield for the wafer group using AM2 in step A and CM2 in step C.

Next, an example of calculation of the two-part confounding degree and the two-part independence degree will be described. In regression tree analysis, the degree of confounding (state of confounding, degree of independence) of each set division state, which was performed to find the most significant explanatory variable for the objective variable, was statistically grasped. Clarify other explanatory variables that are confounded by the explained explanatory variable. Referring to FIG.
The calculation method of the degree of division confounding and the degree of division independence will be described.

(1) Among the explanatory variables, a variable whose confounding degree is to be evaluated is defined as a reference explanatory variable 801.

(2) Each record forms a table in which "L" or "H" is a data value for each explanatory variable. Here, H belongs to a set having a high value of the objective variable when the set is divided into two during regression tree analysis, and L belongs to a set having a low value of the objective variable when the set is divided into two during the regression tree analysis. When the set is divided into two, for each explanatory variable of all records, L,
H is determined.

(3) Based on the reference explanatory variables 801, L and H are used as evaluation values of the degree of coincidence between L and H of each comparative explanatory variable 802.
Let Na be the number of records that match, and N be the total number of records.
The two-part confounding degree DEP is defined as in Expression (5). The range of the two-part confounding degree DEP is -1 to 1, which is 1 if completely confounded, 0 if not confounded at all, and 11 if confounded conversely.

DEP = (2 × Na / N) −1 (5)

Further, based on the two-part confounding degree DEP, the two-part independence degree IND is defined as in equation (6). The range of the two-part independence degree IND is 0 to 1, which is 1 if completely independent and 0 if not completely independent.

IND = 1− | DEP | (6)

(4) The two-part confounding degree DEP and the two-part independence IND are obtained between one reference explanatory variable 801 and another explanatory variable 802, and are used as an evaluation scale between the explanatory variables. Which explanatory variable is used as the reference explanatory variable is arbitrary, but from the viewpoint of its usefulness, it is assumed that the significant difference is large for the target variable in regression tree analysis, especially in the set division at the top hierarchy. Is effective.

(5) By calculating the above-mentioned two-part confounding degree DEP and two-part independence IND, each comparative explanatory variable 8
02 is the reference explanatory variable 80 which belongs to each set of L and H
It is possible to quantitatively evaluate the difference from the one.

By obtaining the degree of two-part confounding and / or the degree of two-part independence, the degree of confounding of explanatory variables can be quantitatively evaluated based on the set of two parts of regression tree analysis. In addition, it is possible to automatically extract another explanatory variable confounded with the explanatory variable having a large significant difference obtained by the regression tree analysis.

The bisection confounding degree can be evaluated for any explanatory variable targeted in the regression tree analysis. However, from the viewpoint of its effectiveness, the explanatory variable (= reference explanatory variable) which is ranked higher in the first division candidate in FIG. Variables, listed in the statistical value list for evaluation) and any other explanatory variables are statistically grasped, and noteworthy explanatory variables that are confounded for explanatory variables with a significant difference are extracted. . Reference explanatory variable 801
An explanatory variable whose degree of confounding is to be analyzed is a comparative explanatory variable 802, and both are selected from the evaluation statistical value list in FIG. The calculation example of the two-part confounding degree and the two-part independence degree is as follows.
This will be described with reference to FIG.

FIG. 8 shows the wafer number 803, the comparative explanatory variable 802, the reference explanatory variable 801, and the yield 804 on the horizontal axis, and the high yield group 811 of the reference explanatory variable on the vertical axis.
Low yield group 812 of reference explanatory variables, calculation formula 813 of two-part confounding degree 813, two-part confounding degree 814, two-part independence 8
15 is shown.

The item used as the reference for comparison among the top candidate items (evaluation statistical value list) in FIG.
Decide as 1. In FIG. 8, ST3 is the reference explanatory variable 80.
It is one. Other explanatory variables are referred to as comparative explanatory variables 802. In FIG. 8, ST1, ST2, and WET2 are comparison explanatory variables 802. Each comparison explanatory variable 802 is compared with the reference explanatory variable 801. ST1, ST2, which are explanatory variables
In ST3 and WET2, "L" of the low yield group is indicated by hatching, and "H" of the high yield group is indicated without hatching.

The reference explanatory variable 801, ST 3, can be divided into a high yield group 811 and a low yield group 812 according to its attribute value. High yield group 8
11 is a set of ten, and the low yield group 812 is also a set of ten.

Next, the number of lots of the high yield group and the low yield group obtained by dividing each of the explanatory variables into two coincides with the same group of the reference explanatory variables, and N is counted.
a. For example, ST1 which is the comparison explanatory variable 802
Is 10 high yield groups included in the high yield group 811 of the reference explanatory variables, and 2 low yield groups included in the low yield group 812 of the reference explanatory variables. That is, the number Na = 10 + 2 = 12 in which the comparison explanatory variable ST1 and the reference explanatory variable ST3 belong to the same group. A formula obtained by substituting this Na into formula (5) is shown in formula 813. Here, the data number N is 20. The calculation result is shown in a two-part confounding degree 814. The value obtained by equation (6) is shown as a two-part independence 815. Two-part confounding degree 814 and two-part independence 815
Are shown below each column in FIG.

There are three basic methods of utilizing the two-part confounding degree and the two-part independence degree. An explanatory variable that has been difficult to discriminate conventionally can be obtained as quantitative information as follows.

(1) Confirming the range of significant explanatory variables Candidates confounded with highly significant candidates are grasped, and these are also determined to be significant explanatory variables. Although there is no particular criterion for the confounding degree, it can be determined by comparing it with the values of other explanatory variables. In addition, when a candidate that does not need to be considered as a technical target comes to the top, a candidate confounded with this candidate can be clarified.
Furthermore, meaningless candidates can be deleted and analyzed again to confirm.

(2) Confirmation of Candidates with High Independence and Its Application The independence of all candidates from other candidates is confirmed, and if there is a candidate with sufficiently high independence from other candidates, the yield by this candidate is determined. It becomes clear that the difference exists independently of the other candidates. Further, similar discriminant tree analysis is performed for each of the candidate divided groups and compared. If the same analysis result is obtained in both cases, it is understood that the reliability of the analysis result is high. Conversely, if the analysis results are different, there is an explanatory variable that affects the yield under compound conditions with the candidate considered independent, or it is influenced by unusual data (due to the small number of data, etc.) Conceivable.

(3) Analysis of Discrimination Tree Regarding Confounding Candidates If a candidate considered important is confounded with a first branch candidate, it is unlikely to appear in a lower branch of the first branch. that time,
The data is divided by another candidate division group having a high degree of independence to perform a discrimination tree analysis, and the results of the discrimination tree analysis under this division group are compared. With similar results, the important candidate cannot be distinguished from the first candidate, but the analysis itself is considered to be highly reliable. Conversely, if the important candidate appears and the result is different, this result should also be considered, and it is necessary to further perform data analysis which can analyze the candidate considered important and the first candidate separately. It is thought that there is.

Next, a regression tree analysis is performed using the apparatus history and the electrical characteristic values as explanatory variables and the wafer yield as an objective variable. The case of obtaining the degree of independence will be described.

FIGS. 9 and 10 show a regression tree diagram and an evaluation statistic list obtained in this embodiment. In FIG. 9, a node n900 is divided into nodes n901 to n914.
FIG. 10 shows evaluation statistical values of the top 12 explanatory variables at the time of the first two divisions. As a result, 12 candidates 1 of the set branch
001 to 1012 are listed.

FIG. 11 shows the first candidate 10 at the top of FIG.
ST1 listed as 01 is used as a reference explanatory variable 1101
The two-part confounding degree 1111 and the two-part independence 1112 when the other explanatory variables in the evaluation statistic value list are comparative explanatory variables 1102 are shown.

FIG. 12 shows the third candidate 1 of the set branch shown in FIG.
ST3 listed as 003 is used as a reference explanatory variable 120.
1 and the two-part confounding degrees 1211 and 211 when the other explanatory variable of the evaluation statistic list is a comparative explanatory variable 1202.
The degree of division independence 1212 is shown.

ST2, ST4, ST5, ST6, and ST1 show that the two-part confounding degree shown in FIG. 11 exceeds 0.75.
0 and WET2, which do not appear in the regression tree diagram of FIG. 9 but may be factors that greatly affect the yield. Conversely, FIG. 12 shows that ST3 has a high degree of two-part independence.

FIG. 12 shows the two-part confounding degree 1211 and the two-part independence degree 1212 with the other 11 explanatory variables using ST3, which is determined to be high in two-part independence in FIG. 11, as a reference explanatory variable. ST3 indicates that the degree of independence from any other explanatory variables is high.

FIGS. 13 and 14 show two of the top 12 explanatory variables determined to have a significant difference in the regression tree analysis of FIG.
The degree of division confounding, the degree of division independence, and the average value thereof are shown, and the relationship between explanatory variables can be grasped at a glance. The lowermost column in FIG. 13 shows the average value 1301 of the two-part confounding degree, and the lowermost column of FIG. 14 shows the average value 1401 of the two-part independence degree.

Next, it was found that the difference in the equipment used in ST3 was effective for the yield independently of the other explanatory variables. : Using S3M2 and S3M3) and the wafer group (good wafer group, using S3M1 and S3M4) by the equipment group in ST3 which is good, and perform regression tree analysis separately. Figure 1 shows the resulting regression tree.
5, shown in FIG.

FIG. 15 is a regression tree diagram showing the result of regression tree analysis based on a group of defective wafers.
506. FIG. 16 is a regression tree diagram showing a regression tree analysis result using a good wafer group, and includes a node n160.
0 to n1606.

The first branch of the defective wafer group in FIG.
This is the same as the case of all wafer groups of FIG. 9. In addition to the fact that the number of defective wafer groups on the top layer of the regression tree diagram of FIG. 9 is as small as n = 39, when the yield is significantly affected by extremely poor wafers compared to the others. It is speculated that this is one of the factors that makes the analysis difficult. In the good wafer group shown in FIG. 16, the factors that were difficult to see due to the defective apparatus in the ST3 process were newly found.

According to the present embodiment, the two-part confounding degree and the two
The degree of confounding of explanatory variables can be grasped more clearly using the degree of split independence, and in combination with regression tree analysis,
It is possible to clarify the explanatory variables to be noted that are confounded with the problem explanatory variables in which the significant difference between the first branches in the regression tree is large.

Further, the accuracy (reliability) of the regression tree analysis and the analysis efficiency are improved by applying regression tree analysis again by applying the grouping of highly independent explanatory variables, and more detailed analysis is possible. Become.

The embodiment of the present invention can be realized by a computer executing a program.
Further, means for supplying the program to the computer, for example, a recording medium such as a CD-ROM in which the program is recorded, or a transmission medium such as the Internet for transmitting the program can be applied as an embodiment of the present invention. The above-described program, recording medium, and transmission medium are included in the scope of the present invention.

It should be noted that each of the above-described embodiments is merely an example of the embodiment for carrying out the present invention, and the technical scope of the present invention should not be interpreted in a limited manner. It is. That is, the present invention can be implemented in various forms without departing from the technical idea or the main features.

The following summarizes various embodiments of the present invention. (Supplementary Note 1) (a) preparing data results of explanatory variables and objective variables; (b) calculating confounding degree and / or independence degree between a plurality of explanatory variables based on the data results; c) performing data mining using the degree of confounding and / or degree of independence. (Supplementary Note 2) The data analysis method according to Supplementary Note 1, wherein the step (b) calculates the confounding degree and / or the independence degree for each set divided into two by regression tree analysis. (Supplementary Note 3) In the step (b), regression tree analysis is performed to select a plurality of explanatory variables that cause a division with a significant difference, and to calculate the degree of confounding and / or independence between the plurality of explanatory variables. 2. The data analysis method according to 2. (Supplementary Note 4) In the step (b), when calculating the degree of confounding and / or the degree of independence between the explanatory variable serving as a reference and other explanatory variables, the explanation in each set divided into two by regression tree analysis is performed. 4. The data analysis method according to claim 3, wherein the degree of confounding and / or the degree of independence is calculated based on the ratio of data coincidence and non-coincidence between variables. (Supplementary Note 5) The step (c) is based on the confounding degree and / or
Alternatively, the data analysis method according to Appendix 4, wherein data mining is performed by selecting explanatory variables based on the degree of independence. (Supplementary Note 6) An arithmetic unit that calculates a degree of confounding and / or independence between a plurality of explanatory variables based on data results of explanatory variables and an objective variable, and performs data mining using the degree of confounding and / or independence. A data analysis device having data mining means. (Supplementary note 7) The data analysis device according to supplementary note 6, wherein the calculation means calculates the confounding degree and / or the independence degree in a set unit divided into two by regression tree analysis. (Supplementary note 8) The supplementary note 7, wherein the calculating means selects a plurality of explanatory variables that cause a division having a significant difference by regression tree analysis, and calculates the degree of confounding and / or independence between the plurality of explanatory variables. Data analysis equipment. (Supplementary Note 9) When calculating the degree of confounding and / or the degree of independence between the explanatory variable serving as a reference and the other explanatory variables, the calculating means may determine whether the explanatory variables in each set are divided into two by regression tree analysis. 9. The data analysis apparatus according to claim 8, wherein the degree of confounding and / or the degree of independence is calculated based on the ratio of coincidence and non-coincidence of the data. (Supplementary note 10) The data analysis device according to supplementary note 9, wherein the data mining unit performs data mining by selecting and explaining explanatory variables based on the confounding degree and / or the independence degree. (Supplementary Note 11) (a) a procedure for preparing data results of explanatory variables and objective variables, (b) a procedure for calculating the degree of confounding and / or independence between a plurality of explanatory variables based on the data results, c) a computer-readable recording medium that records a program for causing a computer to execute the data mining procedure using the confounding degree and / or the independence degree.

[0069]

As described above, according to the present invention, the degree of confounding of explanatory variables can be clearly grasped by calculating the degree of confounding and / or the degree of independence between a plurality of explanatory variables. If a regression tree analysis is performed based on this, it becomes possible to quantitatively evaluate the degree of confounding of explanatory variables based on the result of dividing the set of the regression tree analysis into two, and the problem that the first branch of the regression tree has a significant difference is large. It is possible to clarify the explanatory variable to be noted that is confounded with the explanatory variable to be.

[Brief description of the drawings]

FIG. 1 is a diagram showing a relationship between a lot flow and an abnormal manufacturing apparatus.

FIG. 2 is a diagram showing a yield distribution for each device in a certain process according to the related art.

FIG. 3 is a diagram showing the relationship between the flow of a lot and the confounding of abnormal manufacturing apparatuses.

FIG. 4 is a diagram showing an example of regression tree analysis input data.

FIG. 5 is a diagram illustrating an example of a regression tree.

FIG. 6 is a diagram illustrating an example of an evaluation statistical value list.

FIG. 7 is a diagram showing a relationship among a used manufacturing apparatus, electrical characteristic data, and a yield value.

FIG. 8 is a diagram illustrating a calculation example of a two-part confounding degree and a two-part independence degree.

FIG. 9 is a diagram illustrating an example of a regression tree.

FIG. 10 is a diagram illustrating an example of an evaluation statistical value list.

FIG. 11 is a diagram showing the degree of confounding and the degree of independence between each explanatory variable and the explanatory variable of the first candidate.

FIG. 12 is a diagram illustrating the degree of confounding and the degree of independence between each explanatory variable and an explanatory variable of a third candidate.

FIG. 13 is a diagram showing the degree of confounding among all candidates and the average thereof.

FIG. 14 is a diagram showing the degree of independence of all candidates and the average thereof.

FIG. 15 is a regression tree diagram showing a regression tree analysis result based on a group of defective wafers.

FIG. 16 is a regression tree diagram showing a regression tree analysis result using a good wafer group.

FIG. 17 is a diagram illustrating a configuration of a data analysis device.

[Explanation of symbols]

 101 Normal device 102 Abnormal device 401 Explanatory variable 402 Target variable 411 Used device 412 Electrical characteristic data 413 Yield 801 Reference explanatory variable 802 Comparative explanatory variable 803 Wafer number 804 Yield 811 High yield group of standard explanatory variable 812 Low yield of standard explanatory variable Group 813 Formula for calculating 2-part confounding degree 814 2-part confounding degree 815 2-part independence 1701 Original data group 1702 Database 1703 Data mining unit 1704 Rule file 1705 Analysis tool group 1706 Statistical analysis component 1707 Chart creation component 1708 Decision making unit

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Hidetaka Tsuda 3-2-1 Sakado, Takatsu-ku, Kawasaki-shi, Kanagawa F-term in Fujitsu LSI Technology Co., Ltd. 5B056 BB00 HH00 5B075 ND03 NR12 NR16

Claims (10)

[Claims] (A) preparing a data result of an explanatory variable and an objective variable; and (b) calculating a confounding degree and / or an independence degree between a plurality of explanatory variables based on the data result. (C) performing data mining using the degree of confounding and / or degree of independence. 2. The data analysis method according to claim 1, wherein in the step (b), the confounding degree and / or the independence degree are calculated for each set divided into two by regression tree analysis. 3. In the step (b), a plurality of explanatory variables which cause a division having a significant difference are selected by a regression tree analysis, and the degree of confounding and / or independence between the plurality of explanatory variables is calculated. The data analysis method according to claim 2. 4. The method according to claim 1, wherein the step (b) comprises: calculating a confounding degree and / or an independence degree between a reference explanatory variable and another explanatory variable; 4. The data analysis method according to claim 3, wherein the degree of confounding and / or the degree of independence is calculated based on the ratio of data coincidence and disagreement between explanatory variables. 5. The data analysis method according to claim 4, wherein the step (c) performs data mining by selecting an explanatory variable based on the degree of confounding and / or degree of independence. 6. An arithmetic unit for calculating a degree of confounding and / or independence between a plurality of explanatory variables based on data results of explanatory variables and an objective variable, and performing data mining using the degree of confounding and / or independence. A data analysis device having data mining means for performing the data analysis. 7. The data analysis device according to claim 6, wherein the calculation means calculates the confounding degree and / or the degree of independence for each set divided into two by regression tree analysis. 8. The calculation means selects a plurality of explanatory variables that cause a division having a significant difference by regression tree analysis,
The data analysis device according to claim 7, wherein the degree of confounding and / or the degree of independence between the plurality of explanatory variables is calculated. 9. The calculation means calculates the degree of confounding and / or the degree of independence between a reference explanatory variable and another explanatory variable when the explanatory variable in each set divided into two by regression tree analysis. 9. The data analysis device according to claim 8, wherein the degree of confounding and / or the degree of independence is calculated based on a ratio of data coincidence and non-coincidence between the data. 10. The data analysis apparatus according to claim 9, wherein said data mining means performs data mining by selecting and explaining explanatory variables based on said confounding degree and / or independence degree.