# US 2013/0159226 A1 · Method for Screening Samples for Building Prediction Model and Computer Program Product Thereof > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2013/0159226 A1 - Document ID: US20130159226A1 - Status: Published application - Kind: A1 - Application number: 13/667,039 - Filed: 2012-11-02 - Published: 2013-06-20 - Earliest claimed: 2011-12-20 - Assignees: National Cheng Kung University - Inventors: Fan-Tien Cheng; Wei-Ming Wu - Web page: https://patentgrep.com/patent/US20130159226A1 ## CPC - G06N 3/088: Non-supervised learning, e.g. competitive learning - G03F 7/705: {Modelling or simulating from physical phenomena up to complete wafer processes or whole workflow in wafer productions} - G03F 7/70616: {Monitoring the printed patterns} - G01W 1/10: Devices for predicting weather conditions (computers per se G06; display devices G09) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G03F: PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR; (phototypographic composing devices B41B; photosensitive materials or processes for photographic purposes G03C; electrophotography, sensitive layers or processes therefor G03G) - G01W: METEOROLOGY (radar, sonar, lidar or analogous systems, designed for meteorological use G01S13/95, G01S15/88, G01S17/95) ## Abstract A method for screening samples for building a prediction model and a computer program product thereof are provided. When a set of new sample data is added to a dynamic moving window (DMW), a clustering step is performed with respect to all of the sets of sample data within the window for grouping the sets of sample data with similar properties as one group. If the number of the sets of sample data in the largest group is greater than a predetermined threshold, it means that there are too many sets of sample data with similar properties in the largest group, and the oldest sample data in the largest group can be deleted; if smaller than or equal to a predetermined threshold, it means that the sample data in the largest group are quite unique, and should be kept for building or refreshing the prediction model. ## Claims ### Claim 1 A method for screening samples for building a prediction model, comprising: obtaining a plurality of sets of first sample data sequentially generated with respect to a target to be predicted, the sets of first sample data comprising: a plurality of sets of monitored data; and a plurality of objective data, wherein the objective data are corresponding to the sets of monitored data in a one-to-one manner and are cause-and-result related; performing a clustering step with respect to all of the sets of first sample data for grouping the sets of first sample data with high similarities as one group, thereby forming and obtaining a plurality of first groups; searching for at least one of the first groups having the most number of sets of first sample data, thereby obtaining at least one second group; determining if the number of the at least one second group is greater than or equal to 2, thus obtaining a first determination result; searching for one of the at least one second group having the oldest set of first sample data when the first determination result is yes, thereby obtaining a third group; and determining if the number of sets of first data in the third group is smaller than a predetermined number, thus obtaining a second determination result; determining if the number of sets of first data in the second group is smaller than the predetermined number when the first determination result is no, thus obtaining a third determination result; reserving all of the sets of first sample data for building or refreshing the prediction model when the second determination result or the third determination result is yes, wherein the prediction model is used for predicting a status or behavior of the target; discarding the oldest set of first sample data in the third group and reserving the remaining sets of first sample data for building or refreshing the prediction model when the second determination result is no; and discarding the oldest set of first sample data in the second group and reserving the remaining sets of first sample data for building or refreshing the prediction model when the third determination result is no. ### Claim 2 (depends on 1) The method as claimed in claim 1, wherein the sets of first sample data comprise: a plurality of sets of historical sample data stored in a database; and a set of newly-collected sample data. ### Claim 3 (depends on 1) The method as claimed in claim 1, wherein the sets of first sample data are obtained via a moving window. ### Claim 4 (depends on 1) The method as claimed in claim 1, wherein the target to be predicted is a workpiece processed by a process tool, and the prediction model is a virtual metrology model, wherein the sets of monitored data are a plurality of sets of process data which are generated while a plurality of historical workpieces are processed in the process tool, each of the sets of process data comprising a plurality of process parameters; and the objective data are a plurality of metrology data which are obtained by using a metrology tool to measure the historical workpieces, wherein the metrology data are corresponding to the sets of process data in a one-to-one manner. ### Claim 5 (depends on 1) The method as claimed in claim 1, wherein the clustering step is performed in accordance with an adaptive resonance theory 2 (ART2) algorithm. ### Claim 6 (depends on 1) The method as claimed in claim 1, wherein the clustering step is performed in accordance with a Euclidean distance (ED) algorithm. ### Claim 7 (depends on 1) The method as claimed in claim 1, wherein the clustering step is performed in accordance with a weighted Euclidean Distance (WED) algorithm. ### Claim 8 (depends on 1) The method as claimed in claim 1, wherein the clustering step is performed with respect to the sets of sets of monitored data in the sets of first sample data. ### Claim 9 (depends on 1) The method as claimed in claim 1, wherein the clustering step is performed with respect to the sets of sets of monitored data and the objective data in the sets of first sample data. ### Claim 10 A computer program product stored on a non-transitory tangible computer readable recording medium, which, when executed, performs a method for screening samples for building a prediction model, the method comprising: obtaining a plurality of sets of first sample data sequentially generated with respect to a target to be predicted, the sets of first sample data comprising: a plurality of sets of monitored data; and a plurality of objective data, wherein the objective data are corresponding to the sets of monitored data in a one-to-one manner and are cause-and-result related; performing a clustering step with respect to all of the sets of first sample data for grouping the sets of first sample data with high similarities as one group, thereby forming and obtaining a plurality of first groups; searching for at least one of the first groups having the most number of sets of first sample data, thereby obtaining at least one second group; determining if the number of the at least one second group is greater than or equal to 2, thus obtaining a first determination result; searching for one of the at least one second group having the oldest set of first sample data when the first determination result is yes, thereby obtaining a third group; and determining if the number of sets of first data in the third group is smaller than a predetermined number, thus obtaining a second determination result; determining if the number of sets of first data in the second group is smaller than the predetermined number when the first determination result is no, thus obtaining a third determination result; reserving all of the sets of first sample data for building or refreshing the prediction model when the second determination result or the third determination result is yes, wherein the prediction model is used for predicting a status or behavior of the target; discarding the oldest set of first sample data in the third group and reserving the remaining sets of first sample data for building or refreshing the prediction model when the second determination result is no; and discarding the oldest set of first sample data in the second group and reserving the remaining sets of first sample data for building or refreshing the prediction model when the third determination result is no. ### Claim 11 (depends on 10) The computer program product as claimed in claim 10, wherein the sets of first sample data comprise: a plurality of sets of historical sample data stored in a database; and a set of newly-collected sample data. ### Claim 12 (depends on 10) The computer program product as claimed in claim 10, wherein the sets of first sample data are obtained via a moving window. ### Claim 13 (depends on 10) The computer program product as claimed in claim 10, wherein the target to be predicted is a workpiece processed by a process tool, and the prediction model is a virtual metrology model, wherein the sets of monitored data are a plurality of sets of process data which are generated while a plurality of historical workpieces are processed in the process tool, each of the sets of process data comprising a plurality of process parameters; and the objective data are a plurality of metrology data which are obtained by using a metrology tool to measure the historical workpieces, wherein the metrology data are corresponding to the sets of process data in a one-to-one manner. ### Claim 14 (depends on 10) The computer program product as claimed in claim 10, wherein the clustering step is performed in accordance with an adaptive resonance theory 2 (ART2) algorithm. ### Claim 15 (depends on 10) The computer program product as claimed in claim 10, wherein the clustering step is performed in accordance with a Euclidean distance (ED) algorithm. ### Claim 16 (depends on 10) The computer program product as claimed in claim 10, wherein the clustering step is performed in accordance with a weighted Euclidean Distance (WED) algorithm. ### Claim 17 (depends on 10) The computer program product as claimed in claim 10, wherein the clustering step is performed with respect to the sets of sets of monitored data in the sets of first sample data. ### Claim 18 (depends on 10) The computer program product as claimed in claim 10, wherein the clustering step is performed with respect to the sets of sets of monitored data and the objective data in the sets of first sample data.