Hybridization of ring theory-based evolutionary algorithm and particle swarm optimization to solve class imbalance problem

Abstract Many real-life datasets are imbalanced in nature, which implies that the number of samples present one class (minority class) is exceptionally less compared to found other (majority class). Hence, if we directly fit these a standard classifier for training, then it often overlooks minority while estimating separating hyperplane(s) and as result missclassifies samples. To solve this problem, over years, many researchers have followed different approaches. However selection true representative from majority still considered an open research problem. A better solution problem would be helpful applications like fraud detection, disease prediction text classification. Also, recent studies show needs not only analyzing disproportion between classes, but also difficulties rooted nature data thereby more flexible, self-adaptable, computationally efficient real-time method without loosing much important it. Keeping fact mind, proposed hybrid model constituting Particle Swarm Optimization (PSO), popular swarm intelligence-based meta-heuristic algorithm, Ring Theory (RT)-based Evolutionary Algorithm (RTEA), recently physics-based algorithm. We named algorithm RT-based PSO or short RTPSO. RTPSO can select most takes advantage exploration exploitation phases its parent algorithms strengthening search process. used AdaBoost observe final classification results our model. The effectiveness has been evaluated on 15 having low extreme imbalance ratio. performance with PSO, RTEA undersampling methods. obtained demonstrate superiority state-of-the-art problem-solvers here comparison. source code work available https://github.com/Sayansurya/RTPSO_Class_imbalance .