An Optimized Cost-Free Learning Using ABC-SVM Approach in the Class Imbalance Problem

In this work, cost-free learning (CFL) formally defined in comparison with cost-sensitive learning (CSL). The primary difference between them is that even in the class imbalance problem, a CFL approach provides optimal classification results without requiring any cost information. In point of fact, several CFL approaches exist in the related studies like sampling and some criteria-based approaches. Yet, to our best knowledge none of the existing CFL and CSL approaches is able to process the abstaining classifications properly when no information is given about errors and rejects. Hence based on information theory, here we propose a novel CFL which seeks to maximize normalized mutual information of the targets and the decision outputs of classifiers. With the help of this strategy, we can manage binary or multi-class classifications with or without refraining. Important features are observed from the new strategy. When the degree of class imbalance is changing, this proposed strategy could able to balance the errors and rejects accordingly and automatically. A wrapper paradigm of proposed ABC-SVM (Artificial Bee Colony-SVM) is oriented on the evaluation measure of imbalanced dataset as objective function with respect to feature subset, misclassification cost and intrinsic parameters of SVM. The main goal of cost free ABC-SVM is to directly improve the performance of classification by simultaneously optimizing the best pair of intrinsic parameters, feature subset and misclassification cost parameters. The obtained experimental results on various standard benchmark datasets and real-world data with different ratios of imbalance show that the proposed method is effective in comparison with commonly used sampling techniques.