International Business Machines Corporation
ENSEMBLE VOTING CLASSIFIERS USING ADJUSTED THRESHOLDS

Abstract:

An example system includes a processor to receive training data used to train an ensemble voting classifier. For each classifier in the ensemble voting classifier, the processor can also set a classification score of a positive training item as a threshold. The processor can further adjust a threshold of at least one of the classifiers based on an analysis of a vote contribution of each classifier on the votes on the training data. The threshold of the at least one of the classifiers is adjusted to increase a voting specificity without impacting sensitivity with respect to the training data.