Discrimination likelihood estimate for trained machine learning model

1 . A system comprising:

a memory storing processor-executable program code; and

at least one processing unit to execute the processor-executable program code to cause the system to:

determine a plurality of records, each of the plurality of records associating each of a plurality of features with a respective value, a second feature with a second feature value, and a target feature with a target value;

train a first machine learning model based on the plurality of records to output an inferred target value of the target feature based on first input values of each of the plurality of features;

train a second machine learning model based on the plurality of records to output an inferred second feature value of the second feature based on the second input values of each of the plurality of features;

determine, based on the trained second machine learning model, a first one or more of the plurality of features which are correlated to the second feature;

determine an influence of each of the first one or more features on the trained first machine learning model;

determine a first value indicating a discrimination likelihood of the trained first machine learning model with respect to the second feature based on the determined influences and on the trained second machine learning model;

receive a command to cancel generation of the trained first machine learning model based on the discrimination likelihood; and

in response to the received command, cancel generation of the trained first machine learning model.

2 . A system according to claim 1 , wherein determination of the first one or more of the plurality of features which are correlated to the second one of the first plurality of features comprises determination of one or more of the plurality of features having a greatest influence on the trained second machine learning model.

3 . A system according to claim 2 , the at least one processing unit to execute the processor-executable program code to cause the system to:

train a third machine learning model based on the plurality of records to output an inferred value of a third feature based on third input values of each of the plurality of features except for the third feature and the second feature;

determine, based on the trained third machine learning model, a second one or more of the plurality of features which are correlated to the third feature;

determine an influence of each of the second one or more features on the trained first machine learning model; and

determine a second value indicating a discrimination likelihood of the trained first machine learning model with respect to the third feature based on the determined influences of each of the second one or more features and on the trained third machine learning model; and

determine a composite value associated with the trained first machine learning model based on the first value and the second value.

4 . A system according to claim 3 , the at least one processing unit to execute the processor-executable program code to cause the system to:

determine, based on the trained second machine learning model, a first probability associated with a category of the second feature; and

determine, based on the trained third machine learning model, a second probability associated with a category of the third feature,

wherein determination of the first value is based on the first probability and on the determined influences of each of the first one or more features, and

wherein determination of the second value is based on the second probability and on the determined influences of each of the second one or more features.

5 . A system according to claim 2 , the at least one processing unit to execute the processor-executable program code to cause the system to:

determine, based on the trained second machine learning model, a first probability associated with a category of the second feature,

wherein determination of the first value is based on the determined influences and the first probability.

6 . A system according to claim 2 , wherein training of the second machine learning model comprises:

converting all second feature values of the second feature into either one of a discriminated category and a non-discriminated category.

7 . A system according to claim 6 , the at least one processing unit to execute the processor-executable program code to cause the system to:

determine, based on the trained second machine learning model, a first probability associated with a category of the second feature,

wherein determination of the first value is based on the determined influences and the first probability.

8 . A method comprising:

receiving a plurality of records, each of the plurality of records associating each of a plurality of features with a respective value, a second feature with a second feature value, and a target feature with a target value;

training a first machine learning model based on the plurality of records to output an inferred target value of the target feature based on first input values of each of the plurality of features;

training a second machine learning model based on the plurality of records to output an inferred second feature value of the second feature based on second input values of each of the plurality of features;

determining, based on the trained second machine learning model, a first one or more of the plurality of features which are correlated to the second feature;

determining an influence of each of the first one or more features on the trained first machine learning model;

determining a first discrimination likelihood of the trained first machine learning model with respect to the second feature based on the determined influences and on the trained second machine learning model;

receiving a command to cancel generation of the trained first machine learning model based on the discrimination likelihood; and

in response to the received command, cancelling generation of the trained first machine learning model.

9 . A method according to claim 8 , wherein determining the first one or more of the plurality of features which are correlated to the second one of the first plurality of features comprises determining of one or more of the plurality of features having a greatest influence on the trained second machine learning model.

10 . A method according to claim 9 , further comprising:

training a third machine learning model based on the plurality of records to output a value of a third feature based on third input values of each of the plurality of features except for the third feature and the second feature;

determining, based on the trained third machine learning model, a second one or more of the plurality of features which are correlated to the third feature;

determining an influence of each of the second one or more features to the trained first machine learning model; and

determining a second value indicating a discrimination likelihood of the trained first machine learning model with respect to the third feature based on the determined influences of each of the second one or more features and based on the trained third machine learning model; and

determining a composite discrimination likelihood associated with the trained first machine learning model based on the first value and the second value.

11 . A method according to claim 10 , further comprising:

determining, based on the trained second machine learning model, a first probability associated with a category of the second feature; and

determining, based on the trained third machine learning model, a second probability associated with a category of the third feature,

wherein determining the first value is based on the first probability and the determined influences on the trained second machine learning model, and

wherein determining the second value is based on the determined influences on the trained third machine learning model and the second probability.

12 . A method according to claim 8 , further comprising:

determining, based on the trained second machine learning model, a first probability associated with a category of the second feature,

wherein determining the first value is based on the determined influences and the first probability.

13 . A method according to claim 9 , wherein training of the second machine learning model comprises:

converting all second feature values of the second feature into either one of a discriminated category and a non-discriminated category.

14 . A method according to claim 13 , further comprising:

determining, based on the trained second machine learning model, a first probability associated with a category of the second feature,

wherein determining the first value is based on the determined influences and the first probability.

15 . A non-transitory medium storing executable program code executable by at least one processing unit of a computing system to cause the computing system to:

receive a plurality of records, each of the plurality of records associating each of a plurality of features with a respective value, a second feature with a second feature value, and a target feature with a target value;

train a first machine learning model based on the plurality of records to output an inferred target value of the target feature based on first input values of each of the plurality of features;

train a second machine learning model based on the plurality of records to output an inferred second feature value of the second feature based on second input values of each of the plurality of features;

determine, based on the trained second machine learning model, a first one or more of the plurality of features which are correlated to the second feature;

determine an influence of each of the first one or more features on the trained first machine learning model;

determine a first value indicating a discrimination likelihood of the trained first machine learning model with respect to the second feature based on the determined influences and on the trained second machine learning model;

receive a command to cancel generation of the trained first machine learning model based on the discrimination likelihood; and

in response to the received command, cancel generation of the trained first machine learning model.

16 . A medium according to claim 15 , wherein determination of the first one or more of the plurality of features which are correlated to the second one of the first plurality of features comprises determination of one or more of the plurality of features having a greatest influence on the trained second machine learning model.

17 . A medium according to claim 16 , the program code executable by at least one processing unit of a computing system to cause the computing system to:

train a third machine learning model based on the plurality of records to output an inferred value of a third feature based on third input values of each of the plurality of features except for the third feature and the second feature;

determine, based on the trained third machine learning model, a second one or more of the plurality of features which are correlated to the third feature;

determine an influence of each of the second one or more features on the trained first machine learning model;

determine a second value indicating a discrimination likelihood of the trained first machine learning model with respect to the third feature based on the determined influences of each of the second one or more features and on the trained third machine learning model; and

determine a composite value associated with the trained first machine learning model based on the first value and the second value.

18 . A medium according to claim 17 , the program code executable by at least one processing unit of a computing system to cause the computing system to:

determine, based on the trained second machine learning model, a first probability associated with a category of the second feature; and

determine, based on the trained third machine learning model, a second probability associated with a category of the third feature,

wherein determination of the first value is based on the first probability and on the determined influences of each of the first one or more features, and

wherein determination of the second value is based on the second probability and on the determined influences of each of the second one or more features.

19 . A medium according to claim 16 , the program code executable by at least one processing unit of a computing system to cause the computing system to:

determine, based on the trained second machine learning model, a first probability associated with a category of the second feature,

wherein determination of the first value is based on the determined influences and the first probability.

20 . A medium according to claim 16 , wherein training of the second machine learning model comprises:

converting all second feature values of the second feature into either one of a discriminated category and a non-discriminated category.