Learning with Skewed Class Distributions

摘要

Several aspects may influence the performance achieved by a classifier created by a Machine Learning system. One of these aspects is related to the difference between the numbers of examples belonging to each class. When this difference is large, the learning system may have difficulties to learn the concept related to the minority class. In this work, we discuss several issues related to learning with skewed class distributions, such as the relationship between cost-sensitive learning and class distributions, and the limitations of accuracy and error rate to measure the performance of classifiers. Also, we survey some methods proposed by the Machine Learning community to solve the problem of learning with unbalanced data sets, and discuss some limitations of these methods.