Multi-party prediction using feature contribution values
Techniques are provided for multi-party prediction using feature contribution values. One method comprises obtaining a first set of feature contribution values associated with respective ones of a plurality of machine learning models, wherein each machine learning model is trained using training data of a different party and each feature contribution value indicates a contribution by a corresponding feature to a prediction generated by the associated machine learning model; training an aggregate machine learning model using the obtained first sets of feature contribution values; receiving a second set of feature contribution values generated by applying data of at least one party to at least one machine learning model; and applying the second set of feature contribution values to the trained aggregate machine learning model to obtain a global prediction. Each feature contribution value may correspond to a masked feature, and the feature contribution values may not expose the source data of one party to another party.
1. A method, comprising:
obtaining a first set of feature contribution values associated with respective ones of a plurality of machine learning models, wherein each machine learning model is trained using training data of a different party and wherein each feature contribution value indicates a contribution by a corresponding feature to a prediction generated by the associated machine learning model;
training an aggregate machine learning model by applying the obtained first sets of feature contribution values to the aggregate machine learning model;
receiving at least one second set of feature contribution values generated by applying data of at least one party to one or more of the machine learning models; and
applying the at least one second set of feature contribution values to the trained aggregate machine learning model to obtain a global prediction related to at least some of the data of one or more of the different parties;
wherein the method is performed by at least one processing device comprising a processor coupled to a memory.
2. The method of claim 1 , wherein each of the feature contribution values corresponds to a masked feature.
3. The method of claim 1 , wherein each first set of obtained feature contribution values does not expose the source data of a respective party to a different party.
4. The method of claim 1 , wherein the feature contribution values comprise SHAP values.
5. The method of claim 1 , wherein the training of the aggregate machine learning model further employs at least one label generated by at least one machine learning model.
6. The method of claim 1 , wherein a training data set of a first party is not shared with another party.
7. The method of claim 1 , further comprising employing one or more data access controls to prevent data of a first party from being accessed by another party.
8. The method of claim 1 , wherein the different parties comprise one or more of different data owners, different users and different entities.
9. An apparatus comprising:
at least one processing device comprising a processor coupled to a memory;
the at least one processing device being configured to implement the following steps:
obtaining a first set of feature contribution values associated with respective ones of a plurality of machine learning models, wherein each machine learning model is trained using training data of a different party and wherein each feature contribution value indicates a contribution by a corresponding feature to a prediction generated by the associated machine learning model;
training an aggregate machine learning model by applying the obtained first sets of feature contribution values to the aggregate machine learning model;
receiving at least one second set of feature contribution values generated by applying data of at least one party to one or more of the machine learning models; and
applying the at least one second set of feature contribution values to the trained aggregate machine learning model to obtain a global prediction related to at least some of the data of one or more of the different parties.
10. The apparatus of claim 9 , wherein each first set of obtained feature contribution values does not expose the source data of a respective party to a different party.
11. The apparatus of claim 9 , wherein the feature contribution values comprise SHAP values.
12. The apparatus of claim 9 , wherein the training of the aggregate machine learning model further employs at least one label generated by at least one machine learning model.
13. The apparatus of claim 9 , further comprising employing one or more data access controls to prevent data of a first party from being accessed by another party.
14. The apparatus of claim 9 , wherein the different parties comprise one or more of different data owners, different users and different entities.
15. A non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes the at least one processing device to perform the following steps:
obtaining a first set of feature contribution values associated with respective ones of a plurality of machine learning models, wherein each machine learning model is trained using training data of a different party and wherein each feature contribution value indicates a contribution by a corresponding feature to a prediction generated by the associated machine learning model;
training an aggregate machine learning model by applying the obtained first sets of feature contribution values to the aggregate machine learning model;
receiving at least one second set of feature contribution values generated by applying data of at least one party to one or more of the machine learning models; and
applying the at least one second set of feature contribution values to the trained aggregate machine learning model to obtain a global prediction related to at least some of the data of one or more of the different parties.
16. The non-transitory processor-readable storage medium of claim 15 , wherein each first set of obtained feature contribution values does not expose the source data of a respective party to a different party.
17. The non-transitory processor-readable storage medium of claim 15 , wherein the feature contribution values comprise SHAP values.
18. The non-transitory processor-readable storage medium of claim 15 , wherein the training of the aggregate machine learning model further employs at least one label generated by at least one machine learning model.
19. The non-transitory processor-readable storage medium of claim 15 , further comprising employing one or more data access controls to prevent data of a first party from being accessed by another party.
20. The non-transitory processor-readable storage medium of claim 15 , wherein the different parties comprise one or more of different data owners, different users and different entities.