Effects of Different Base Classifiers to Learn++ Family Algorithms for Concept Drifting and Imbalanced Pattern Classification Problems

This paper is about evaluation of base classifiers in imbalanced and non-stationary environment.

Download

Abstract

Machine learning with concept drifting attracts a lot of attention in recent years. However, there are only a few works on concept drift learning with imbalanced data. The Learn++.NSE, the Learn++.NIE, and the Learn++.CDS from the Learn++ family are three state-of-the-art learning algorithms designed to deal with machine learning with concept drifting. In this work, we firstly give a brief introduction to the Learn++ family algorithms. Then, the Learn++.NIE is tested using five different types of base classifiers on several benchmarking imbalanced drifting datasets to evaluate the effects of classifiers to the Learn++.NIE. It is found that the MLPNN and the CART perform better than others for imbalanced pattern classification problems using the Learn++.NIE. Then, the three Learn++ algorithms are compared using both the MLPNN and the CART as base classifiers. Using the same base classifiers, the Learn++.NIE and the Learn++.CDS yield better recall in minority class. The Learn++.NSE yields better overall accuracies in experiments of some datasets but yields the worst recall in minority class.