METHOD FOR BALANCING DATASETS OF MULTI-CLASS INSTANCE DATA

摘要

This disclosure describes a method for balancing datasets of instances inwhich each instancemay be labelled by a sequence, plurality or distribution of class labels. Thedisclosure includesperforming stochastic under-sampling (removal of dataset instances) andoversampling(replication of dataset instances) based on the distribution of classes ineach instance, tominimize the ratio between the sizes of the minority class (i.e. classlabelling the fewest framesacross all instances) and the majority class (i.e. class labelling the mostframes across allinstances).