Analysis of sampling based classification techniques to overcome class imbalancing
Anjana Gosain, Anju Saha, Navneet Pratap Singh · International Conference on Computing for Sustainable Global Development · 2016
Classifiers are built for analysis through the Knowledge Discovery Process. Usually classifiers are developed assuming that the different classes on which it is built consists of an equal number of instances. But in real life applications, class(es) under consideration usually consists of very less number of instances as compared to other classes involved in the development of classification models (classifiers). Analyzing such an imbalanced data remains a major challenge. In this research paper, we have proposed taxonomy for the methods used for handling class imbalance problem in classification and set out to compare heuristic — based sampling algorithms used for the purpose of analyzing imbalanced datasets.