Frequent Relational Smote For Class Imbalance An Experimental Approach

Abstract

Over the year, our society has been trying to implement such kind of computer which is intelligent, efficient, and perfect. Computer intelligence is based upon previous data analysis. Class imbalance is one of the significant difficulties in the machine learning field, which degrade the performance of analysis and decision-making ability. In this paper we propose FR-SMOTE to mitigate the class imbalance problem. It aids in the improvement of learning performance in the context of an imbalanced learning environment. Our result analysis shows a significant improvement over the existing state-of-art, where we can see the achievement of 86% true positive rate over the imbalanced classes in machine learning algorithm. In this paper we apply several machine learning algorithm like SVM, KNN, Random Forest in highly unbalanced data(WOS), after applying RUS, OS-SMOTE and implemented FR-SMOTE in the base of accuracy, Precision, Recall and F1 Score.