This book investigates the nature of imbalanced data sets and looks at two external methods, which can increase a learner’s performance on under represented classes. Both techniques artificially balance the training data; one by randomly re-sampling examples of the under represented class and adding them to the training set, the other by randomly removing examples of the over represented class from the training set. A combination scheme is then presented. The approach is one in which multiple classifiers are arranged in a hierarchical structure according to their sampling techniques. The architecture consists of two experts, one that boosts performance by combining classifiers that re-sample training data at different rates, the other by combining classifiers that remove data from the training data at different rates. Using the F-measure, which combines precision and recall as a performance statistic, the combination scheme is shown to be effective at learning from severely imbalanced...
