Financial Fraud Detection using Bio-Inspired Key Optimization and Machine Learning Technique

Abstract

Financial fraud (FF) is a serious issue, a big concern for financial institutes and requires more attention. Financial frauds are a variety of theft or credit card stealing that occurs when someone takes money with the purpose of personal gain. In the current scenario, a highly accurate system is required to detect financial fraud and prevent financial loss. In this paper, new FF detection model has been proposed for accurate detect financial frauds using one of the Bio-inspired optimization algorithms as particle swarm optimization (PSO), feature selection (CFS) method and adaptive three machine learning techniques (as SVM, KNN, and AdaboostM1). The proposed model is capable to reduce the false-positive rate, training & testing time and enhanced accuracy. Therefore, the proposed PSOS algorithm applied as a features optimization method for selecting the best subset features from the highly imbalanced dataset to increased detection accuracy. The experiment results point out that the PSOS algorithm is the most excellent features optimization method and selected 7 features out of 15 features. The proposed model has been enhanced accuracy (from 82.90% to 85.51%), TPR (from 80.50 % to 90.20 %), FPR (from 0.151% to 0.191%), and decreased Error Rate (14.92 % to 17.10%) compared to RIG method. The findings of this study clearly illustrate that the Bioinspired algorithm (PSOS) gives more appropriate results compared with the non-Bio-inspired algorithm (RIG).