| CPC G06N 20/20 (2019.01) | 20 Claims |
|
1. A computer implemented method, comprising:
training an ensemble model using a set of training data to generate a trained ensemble model, wherein the trained ensemble model comprises a plurality of component models;
processing, by the trained ensemble model, a set of non-training data to generate a first set of output, wherein the plurality of component models processes the set of non-training data to generate a plurality of component scores for the plurality of component models, and wherein the first set of output is generated from the plurality of component scores, wherein the trained ensemble model executes in a first computing time being an aggregate of executing each of the plurality of component models;
creating an abridged model comprising a set of component models that is a portion of the plurality of component models, wherein the abridged model executes in a second computing time that is less than the first computing time based on executing only the portion of the plurality of component models, wherein creating the abridged model comprises:
testing each subset of a plurality of subsets of the plurality of component models, wherein testing a subset comprises:
determining, for the subset, a deviation between a first set of component scores of the set of component models, and
calculating, for the subset, an accuracy of the subset from the first set of component scores of the subset,
ranking the plurality of subsets of the component models based on the deviation and according to having the accuracy satisfying an accuracy threshold, and
selecting the portion of the plurality of component models for the abridged model using the ranking,
wherein the first set of component scores is in the plurality of component scores;
receiving an input;
processing the input by the abridged model to generate a second set of component scores for the set of component models and an abridged score for the abridged model,
wherein the abridged score is an aggregation of the second set of component scores;
processing the second set of component scores to determine a standard deviation of the second set of component scores;
comparing the standard deviation of the second set of component scores with a deviation threshold to select one of the abridged score and an ensemble score of the ensemble model as an output; and
presenting the output.
|