초록 ▼

A method for data mining information obtained in an integrated circuit fabrication factory (“fab”) that includes steps of: (a) gathering data from the fab from one or more of systems, tools, and databases that produce data in the fab or collect data from the fab; (b) formatting the data and storing

A method for data mining information obtained in an integrated circuit fabrication factory (“fab”) that includes steps of: (a) gathering data from the fab from one or more of systems, tools, and databases that produce data in the fab or collect data from the fab; (b) formatting the data and storing the formatted data in a source database; (c) extracting portions of the data for use in data mining in accordance with a user specified configuration file; (d) data mining the extracted portions of data in response to a user specified analysis configuration file; (e) storing results of data mining in a results database; and (f) providing access to the results.

대표청구항 ▼

1. A method of data mining information obtained in an integrated circuit fabrication factory (“fab”), comprising the steps of:gathering data from the fab from one or more of systems, tools, and databases that produce data in the fab or collect data from the fab; formatting the data and storing the f

1. A method of data mining information obtained in an integrated circuit fabrication factory (“fab”), comprising the steps of:gathering data from the fab from one or more of systems, tools, and databases that produce data in the fab or collect data from the fab; formatting the data and storing the formatted data in a source database; extracting portions of the data for use in data mining in accordance with a user specified configuration file; data mining the extracted portions of data in response to a user specified analysis configuration file; and storing results of data mining in a results database; and providing access to the results; wherein the step of extracting includes obtaining a hypercube definition using the configuration file, using the hypercube definition to create a vector cache definition, and creating a vector cache of information. 2. The method of claim 1 wherein:the step of storing further includes extracting data from the source database and storing the extracted data in a hybrid database that comprises a relational database component and a file system component; and the step of creating a vector cache of information includes: (a) retrieving a list of files and data elements identified by the vector cache definition from the relational database component using hash-index keys; (b) retrieving the files from the file system component; and (c) populating the vector cache with data elements identified in the vector cache definition. 3. The method of claim 2 wherein the step of extracting further includes generating hypercubes from the vector cache information using the hypercube definition.4. A method for data mining information obtained in an integrated circuit fabrication factory, comprising the steps of:gathering data from the fab from one or more of systems, tools, and databases that produce data in the fab or collect data from the fab; formatting the data and storing the formatted data in a source database; extracting portions of the data for use in data mining in accordance with a user specified configuration file; data mining the extracted portions of data in response to a user specified analysis configuration file; storing results of data mining in a results database; and providing access to the results; wherein the step of data mining includes Self Organized Map data mining to form clusters, Map Matching analysis on output from the Self Organized Map data mining to perform cluster matching, Rules Induction data mining on output from the Self Organized Map data mining analysis a rules explanations of clusters, correlating categorical data to numerical data on output from the Rules Induction data mining; and correlating numerical data to categorical data on output from the Map Matching data mining. 5. The method of claim 4 wherein the step of Self Organized Map data mining automatically organizes data, and identifies separate and dominant data clusters that represent different “fab issues” within a data set, and Map Matching analysis describes an “of interest” variable in terms of any historical data that is known to be impacting the behavior of the “of interest” variable on a cluster by cluster basis.6. A method of data mining information obtained in a semiconductor fabrication factory, wherein the factory includes one or more process tools or measurement tools for fabricating or testing semiconductor circuits on substrates, comprising the sequential steps of:reading, from one or more databases, data gathered from the tools, wherein the data includes one or more of measurements and fabrication process parameters; performing a Self Ordered Map neural network analysis of the data to form a Self Ordered Map of the data so that the Self Ordered Map includes one or more clusters of similar data; performing a Rule Induction analysis of at least one of the clusters so as to output one or more hypotheses that explain the at least one cluster; and performing a data mining analysis on the output from the Rule Induction analysis so as to identify a measurement or a process tool setting that is correlated with the one or more hypotheses. 7. A method according to claim 6, wherein the step of performing data mining analysis on the output from the Rule Induction includes:performing a Reverse MahaCu analysis. 8. A method according to claim 6, wherein the step of performing data mining analysis on the output from the Rule Induction includes:performing a data mining algorithm that correlates categorical data or attributes data to numerical data. 9. A method according to claim 6, wherein the step of performing a Self Ordered Map neural network analysis includes:forming a Self Ordered Map that includes clusters of related yield data. 10. A method of data mining information obtained in a semiconductor fabrication factory, wherein the factory includes one or more process tools or measurement tools for fabricating or testing semiconductor circuits on substrates, comprising the steps of:reading, from one or more databases, data gathered from the tools, wherein the data includes a series of data records, wherein each data record includes a value for each of a number of variables, and wherein each variable is a measurement or a fabrication process parameter; performing a Self Ordered Map neural network analysis of the data to create a Self Ordered Map having a layer corresponding to each variable, wherein each layer includes an array of cells such that each cell is characterized by a value, and wherein the layer corresponding to one of the variables is characterized by at least one cluster of cells having values that are either greater than a high threshold value or less than a low threshold value; and performing a Map Matching analysis of one of the clusters so as to output an identification of one or more other variables having a statistical impact on said one variable. 11. A method according to claim 10, wherein said one variable is a yield parameter.12. A method according to claim 10, wherein the step of performing the Map Matching analysis comprises the steps of:identifying said one cluster of cells as the source cluster; if the cells of said source cluster have values greater than said high threshold, then identifying as target cluster cells all cells in the same layer as the source cluster having values less than said low threshold; otherwise, if the cells of said source cluster have values less than said low threshold, then identifying as target cluster cells all cells in the same layer as the source cluster having values greater than said high threshold; projecting the source cluster cells and the target cluster cells onto respective cells of the layers of the Self Ordered Map corresponding to variables other than said one variable; and ranking each of said other variables according to the degree of difference between the source cluster cells and the target cluster cells in the layer of the Self Ordered Map corresponding to that variable. 13. A method according to claim 12, wherein the step of ranking each of said other variables comprises determining the distance between a centroid of the source cluster cells and a centroid of the target cluster cells in the layer of the Self Ordered Map corresponding to that variable.14. A method according to claim 12, wherein the step of ranking each of said other variables comprises determining a number of data records corresponding to the source cluster cells and a number of data records corresponding to the target cluster cells in the layer of the Self Ordered Map corresponding to that variable.15. A method according to claim 12, wherein the step of ranking each of said other variables comprises determining the mean of the values of the cells at the perimeter of each cluster as compared to the mean of the values of the cells surrounding that cluster in the layer of the Self Ordered Map corresponding to that variable.16. A method according to claim 12, wherein the step of ranking each of said other variables comprises the steps of:identifying as source data records a number of the data records that correspond to the source cluster cells; identifying as target data records a number of the data records that correspond to the target cluster cells; determining a T-test score for a first population consisting of the source data records and a second population consisting of the target data records; determining a mean of the value of that variable for the source data records; determining a reducing ratio for that variable as the absolute value of a first value divided by a second value, wherein the first value is the difference between (i) the number of target data records for which the value of that variable is less than said mean and (ii) the number of target data records for which the value of that variable is greater than said mean, and wherein the second value is the number of target data records; and determining a modified T-test score for that variable as the reducing ratio for that variable multiplied by the T-test score. 17. A method according to claim 12, further comprising the steps of:changing at least one of the low threshold value and the high threshold value; and repeating the step of performing the Map Matching analysis. 18. A method according to claim 12, further comprising the steps of:identifying a subset of said data records for which the value of said one variable is less than the low threshold value or greater than the high threshold value; and performing additional data mining analysis of said subset of data records so as to exclude from the additional data mining analysis all data records for which the value of said one variable is between the low threshold value and the high threshold value. 19. A method according to claim 18, wherein the additional data mining analysis comprises Rule Induction analysis to generate rules that explain the interaction of the variables.20. A method according to claim 10, further comprising the step of:performing a data mining analysis on the output from the Map Matching analysis so as to identify categorical data that is correlated said output. 21. A method according to claim 20, wherein the step of performing data mining analysis on the output from the Map Matching includes:performing a MahaCu analysis. 22. A method according to claim 20, wherein the step of performing data mining analysis on the output from the Map Matching includes:identifying a process tool that is correlated with said output. 23. A method of data mining information obtained in a semiconductor fabrication factory, wherein the factory includes one or more process tools or measurement tools for fabricating or testing semiconductor circuits on substrates, comprising the steps of:reading, from one or more databases, data gathered from the tools, wherein the data includes a series of data records, wherein each data record includes a value for each of a number of variables, and wherein each variable is a measurement or a fabrication process parameter; performing a Self Ordered Map neural network analysis of the data to create a Self Ordered Map having a layer corresponding to each variable, wherein each layer includes an array of cells such that each cell is characterized by a value, and wherein the layer corresponding to one of the variables is characterized by at least one cluster of cells having values that are either greater than a high threshold value or less than a low threshold value; and performing additional data mining analysis of a subset of the data records, wherein the subset excludes all data records for which the value of said one variable is between the low threshold value and the high threshold value. 24. A method according to claim 23, wherein the additional data mining analysis comprises Rule Induction analysis to generate rules that explain the interaction of the variables.