| CPC G06N 20/20 (2019.01) [G06F 9/505 (2013.01); G06F 9/5083 (2013.01); G06F 18/214 (2023.01); G06F 18/2185 (2023.01); G06F 18/24323 (2023.01)] | 18 Claims |
|
1. A method comprising:
training, by a computer system, a tree-based ensemble classifier using a training data set, the training resulting in a plurality of decision trees within the tree-based ensemble classifier;
receiving, by the tree-based ensemble classifier, a request to classify a first query data instance using the tree-based ensemble classifier; and
in response to the request:
selecting, by the tree-based ensemble classifier, a first subset of the plurality of decision trees for processing the first query data instance, the first subset excluding one or more decision trees in the plurality of decision trees;
providing, by the tree-based ensemble classifier, the first query data instance as input to each decision tree in the first subset;
combining, by the tree-based ensemble classifier, per-tree classifications generated by the decision trees in the first subset to generate a subset classification;
determining, by the tree-based ensemble classifier, whether a confidence level associated with the subset classification meets or exceeds a threshold;
upon determining that the confidence level meets or exceeds the threshold, outputting the subset classification as a final classification result for the first query data instance; and
upon determining that the confidence level does not meet or exceed the threshold, repeating the selecting, the providing, the combining, and the determining.
|