Learning from imbalanced data using methods of sample selection
Ikram Chairi, Souad Alaoui, Abdelouahid Lyhyaoui · 2012
The majority of Machine Learning (ML) habitually assume that the training sets used for learning are balanced. However, in real world application this hypothesis is not always true. The problem of between-class imbalance is a challenge that has attracted growing attention from both academia and industry because of his critical influence on the performance of machine learning. Many solutions are proposed to resolve this problem: Generally, the common practice for dealing with imbalanced data sets is to rebalance them artificially by using sampling methods. On the other hand, researches show that Sample Selection (SS) methods help to improve the accuracy during the learning process. The main idea of our work is to apply a technique of Sample Selection on the majority class to achieve an undersampling for the imbalanced data. This procedure consent to deal with the imbalance problem and to improve the performance of learning.