Rational decision-making tool for semiconductor processes

1. A method for generating a robust predictive model for a target feature in a semiconductor process, comprising:

identifying a plurality of minimum performance criteria for the target feature;

generating a plurality of different predictive models each configured to predict the target feature;

for each of the plurality of predictive models that meet the plurality of minimum performance criteria for the target feature, comparing a plurality of true positive results for the target feature with a plurality of false positive results for the target feature;

based on defined selection criteria for the comparison of true positive results with false positive results, selecting one of the plurality of predictive models that meet the plurality of minimum performance criteria; and

deploying the selected predictive model into the semiconductor process.

2. The method of claim 1 , the selecting step further comprising:

displaying the true positive results and the false positive results to a user in a user interface; and

manually selecting, by the user, one of the plurality of predictive models based on a preference for a lower number of false positive results with a user-specified constraint on the true positive results.

3. The method of claim 1 , the selecting step further comprising:

displaying the true positive results and the false positive results to a user in a user interface; and

manually selecting, by the user, one of the plurality of predictive models based on a preference for a higher number of true positive results with a user-specified constraint on the false positive results.

4. The method of claim 1 , the selecting step further comprising:

automatically selecting one of the plurality of predictive models and setting a plurality of respective thresholds for a plurality of key parameters for the selected model based on user utility and/or preferences, the key parameters associated with the target feature.

5. The method of claim 4 , further comprising:

displaying the true positive results and the false positive results to a user in a user interface;

receiving user input from the user through the user interface to establish the user utility and/or preferences for the plurality of key parameters; and

automatically selecting one of the plurality of predictive models based on the user input.

6. The method of claim 5 , further comprising:

generating and displaying a survey to the user for guiding the user through a plurality of trade-offs associated with the plurality of key parameters of the target feature; and

receiving and storing responses to the survey from the user;

extracting user utility and/or preferences for the plurality of trade-offs associated with key parameter from the user responses; and

selecting one of the plurality of predictive models based on the customer utility and/or preferences.

7. The method of claim 6 , the generating step further comprising:

configuring a mathematical function to evaluate and present for display the plurality of trade-offs based on the user utility and/or preferences.

8. The method of claim 6 , the generating step further comprising:

configuring a plurality of different mathematical function to separately evaluate and present for display the plurality of trade-offs based on the user utility and/or preferences.

9. The method of claim 8 , further comprising:

configuring at least a utility function and a preference function to separately evaluate and present for display the plurality of trade-offs.

10. A method for generating a robust predictive model for a target feature in a semiconductor process, comprising:

identifying a plurality of minimum performance criteria for the target feature;

generating a plurality of different predictive models each configured to predict the target feature;

performing a comparative analysis of results from each of the plurality of predictive models that meet the plurality of minimum performance criteria for the target feature;

displaying the comparative analysis of results in a user interface;

receiving user input from the user interface to establish customer preferences and/or customer utility for each of a plurality of key parameters associated with the target feature;

selecting one of the plurality of predictive models that meet the plurality of minimum performance criteria based on the customer preferences and/or customer utility for each of the plurality of key parameters; and

deploying the selected predictive model into the semiconductor process.

11. The method of claim 10 , further comprising:

performing the comparative analysis using a binary function.

12. The method of claim 11 , further comprising:

displaying results of the binary function in the user interface as trade-offs for each of the plurality of key parameters.

13. The method of claim 11 , the selection step further comprising:

selecting one of the plurality of predictive models based on a customer preference for a lower number of false positive results with a user-specified constraint on true positive results.

14. The method of claim 11 , the selection step further comprising:

selecting one of the plurality of predictive models based on a customer preference for a higher number of true positive results with a user-specified constraint on false positive results.

15. The method of claim 10 , further comprising:

performing the comparative analysis using a multiclass function.

16. The method of claim 15 , further comprising:

displaying results of the multiclass function in the user interface as a plurality of categories for each of the key parameters; and

selecting one of the plurality of predictive models based on customer preferences within each category.

17. The method of claim 10 , further comprising:

performing the comparative analysis using a statistical function.

18. The method of claim 10 , further comprising:

performing the comparative analysis using a regression function.

19. The method of claim 10 , the selecting step further comprising:

manually selecting one of the plurality of predictive models based on the customer preferences and/or customer utility for each of the plurality of key parameters.

20. The method of claim 10 , the selecting step further comprising:

automatically selecting one of the plurality of predictive models based on the customer preferences and/or customer utility for each of the plurality of key parameters.

21. The method of claim 10 , the step of receiving user input further comprising:

generating and displaying a survey in the user interface for guiding a user through a series of questions designed to obtain customer preferences and/or customer utility associated with the plurality of key parameters of the target feature;

receiving and storing responses to the survey from the user;

extracting customer preferences and/or customer utility for each of the key parameters from the user responses; and

selecting one of the plurality of predictive models based on the customer preferences and/or customer utility.

22. A robust predictive model for predicting a target feature in a semiconductor process including a processor and non-transitory storage for program instruction, the program instructions configured for causing the processor to:

identify a plurality of minimum performance criteria for the target feature;

generate a plurality of different predictive models each configured to predict the target feature;

perform a comparative analysis of results from each of the plurality of predictive models that meet the plurality of minimum performance criteria for the target feature;

display the comparative analysis of results in a user interface;

receive user input through the user interface to establish customer preferences and/or customer utility for each of a plurality of key parameters associated with the target feature; and

select one of the plurality of predictive models that meet the plurality of minimum performance criteria based on the customer preferences and/or customer utility for each of the plurality of key parameters.

23. The robust predictive model of claim 22 , the program instructions further configured for causing the processor to:

generate and display a survey in the user interface for guiding a user through a a series of questions designed to obtain customer preferences and/or customer utility associated with the plurality of key parameters of the target feature;

receive and store responses to the survey from the user; and

extract customer preferences and/or customer utility for each of the key parameters from the user responses.

24. The robust predictive model of claim 22 , the program instructions further configured for causing the processor to:

perform the comparative analysis using a binary function; and

display results of the binary function in the user interface as trade-offs for each of the plurality of key parameters.

25. The robust predictive model of claim 22 , the program instructions further configured for causing the processor to:

perform the comparative analysis using a multiclass function.

26. The robust predictive model of claim 22 , the program instructions further configured for causing the processor to:

perform the comparative analysis using a statistical function.

27. The robust predictive model of claim 22 , the program instructions further configured for causing the processor to:

perform the comparative analysis using a regression function.