Categorical input machine learning models
There is a need for more effective and efficient predictive data analysis based at least in part on categorical input data. This need can be addressed by, for example, solutions for performing predictive data analysis that utilize at least one of categorical level merging, mutual-information-based feature filtering, feature-correlation-based feature filtering to generate training data feature value arrangements, as well as training and using categorical input machine learning models trained using the training data feature value arrangements.
1. A computer-implemented method comprising:
receiving, by one or more processors, a request comprising a prediction input data entry;
inputting, by the one or more processors, the prediction input data entry to a categorical input machine learning model to receive an inferred prediction for the prediction input data entry as an output of the categorical input machine learning model, wherein:
(i) the categorical input machine learning model defines a plurality of training data feature value arrangements, and
(ii) the inferred prediction is generated based on a matching subset of the plurality of training data feature value arrangements that correspond to the prediction input data entry;
identifying, by the one or more processors, a selected training data feature value arrangement from the matching subset of the plurality of training data feature value arrangements that correspond to the prediction input data entry based on a comparison between (a) a mean-wide cross-entropy measure corresponding to the selected training data feature value arrangement and (b) a plurality of mean-wide cross-entropy measures respectively corresponding to the plurality of training data feature value arrangements within the matching subset; and
in response to receipt of the request, providing, by the one or more processors, the inferred prediction with explanatory metadata descriptive of the selected training data feature value arrangement.
2. The computer-implemented method of claim 1 , further comprising:
generating the plurality of mean-wide cross-entropy measures that respectively correspond to the plurality of training data feature value arrangements within the matching subset.
3. The computer-implemented method of claim 1 , further comprising:
generating a plurality of weighted point probability estimation values that respectively correspond to the plurality of training data feature value arrangements within the matching subset; and
generating the inferred prediction by combining the plurality of weighted point probability estimation values.
4. The computer-implemented method of claim 1 , wherein:
the inferred prediction is determined based at least in part on the selected training data feature value arrangement of the plurality of training data feature value arrangements within the matching subset, and
generating the explanatory metadata is performed based at least in part on the selected training data feature value arrangement.
5. The computer-implemented method of claim 1 , wherein the explanatory metadata comprises a model complexity measure for the selected training data feature value arrangement.
6. The computer-implemented method of claim 1 , further comprising enabling display of a prediction output user interface.
7. The computer-implemented method of claim 6 , wherein the prediction output user interface is configured to display the inferred prediction and the explanatory metadata.
8. The computer-implemented method of claim 1 , wherein the selected training data feature value arrangement is identified based on the plurality of mean-wide cross-entropy measures respectively corresponding to the matching subset of the plurality of training data feature value arrangements.
9. A system comprising memory and one or more processors communicatively coupled to the memory, the one or more processors configured to:
receive a request comprising a prediction input data entry;
input the prediction input data entry to a categorical input machine learning model to receive an inferred prediction for the prediction input data entry as an output of the categorical input machine learning model, wherein:
(i) the categorical input machine learning model defines a plurality of training data feature value arrangements, and
(ii) the inferred prediction is generated based on a matching subset of the plurality of training data feature value arrangements that correspond to the prediction input data entry;
identify a selected training data feature value arrangement from the matching subset of the plurality of training data feature value arrangements that correspond to the prediction input data entry based on a comparison between (a) a mean-wide cross-entropy measure corresponding to the selected training data feature value arrangement and (b) a plurality of mean-wide cross-entropy measures respectively corresponding to the plurality of training data feature value arrangements within the matching subset; and
in response to receipt of the request, provide the inferred prediction with explanatory metadata descriptive of the selected training data feature value arrangement.
10. The system of claim 9 , wherein the one or more processors are further configured to:
generate the plurality of mean-wide cross-entropy measures that respectively correspond to the plurality of training data feature value arrangements within the matching subset.
11. The system of claim 9 , wherein the one or more processors are further configured to:
generate a plurality of weighted point probability estimation values that respectively correspond to the plurality of training data feature value arrangements within the matching subset; and
generate the inferred prediction by combining the plurality of weighted point probability estimation values.
12. The system of claim 9 , wherein:
the inferred prediction is determined based at least in part on the selected training data feature value arrangement of the plurality of training data feature value arrangements within the matching subset, and
the one or more processors are further configured to generate the explanatory metadata based at least in part on the selected training data feature value arrangement.
13. The system of claim 9 , wherein the explanatory metadata comprises a model complexity measure for the selected training data feature value arrangement.
14. The system of claim 9 , wherein the one or more processors are further configured to enable display of a prediction output user interface.
15. The system of claim 14 , wherein the prediction output user interface is configured to display the inferred prediction and the explanatory metadata.
16. One or more non-transitory computer-readable storage media including instructions that, when executed by one or more processors, cause the one or more processors to:
receive a request comprising a prediction input data entry;
input the prediction input data entry to a categorical input machine learning model to receive an inferred prediction for the prediction input data entry as an output of the categorical input machine learning model, wherein:
(i) the categorical input machine learning model defines a plurality of training data feature value arrangements, and
(ii) the inferred prediction is generated based on a matching subset of the plurality of training data feature value arrangements that correspond to the prediction input data entry;
identify a selected training data feature value arrangement from the matching subset of the plurality of training data feature value arrangements that correspond to the prediction input data entry based on a comparison between (a) a mean-wide cross-entropy measure corresponding to the selected training data feature value arrangement and (b) a plurality of mean-wide cross-entropy measures respectively corresponding to the plurality of training data feature value arrangements within the matching subset; and
in response to receipt of the request, provide the inferred prediction with explanatory metadata descriptive of the selected training data feature value arrangement.
17. The one or more non-transitory computer-readable storage media of claim 16 , wherein the one or more processors are further configured to:
generate the plurality of mean-wide cross-entropy measures that respectively correspond to the plurality of training data feature value arrangements within the matching subset.
18. The one or more non-transitory computer-readable storage media of claim 16 , wherein the one or more processors are further configured to:
generate a plurality of weighted point probability estimation values that respectively correspond to the plurality of training data feature value arrangements within the matching subset; and
generate the inferred prediction by combining the plurality of weighted point probability estimation values.
19. The one or more non-transitory computer-readable storage media of claim 16 , wherein:
the inferred prediction is determined based at least in part on the selected training data feature value arrangement of the plurality of training data feature value arrangements within the matching subset, and
the one or more processors are further configured to generate the explanatory metadata based at least in part on the selected training data feature value arrangement.
20. The one or more non-transitory computer-readable storage media of claim 16 , wherein the explanatory metadata comprises a model complexity measure for the selected training data feature value arrangement.