Credit Card Fraud Detection using Ensemble Methods in Machine Learning

Abstract

Credit cards are being used vastly by many people nowadays. As the usage is increasing, the frauds are also increasing. Due to fraudulent transactions, many common people are suffering a lot. Hence, the need to find out whether a given transaction is legit or not is a must these days. The advancement of technology also brings with it an increase in risks. Hence, a good technique for detecting fraudulent transactions is a requisite. This paper proposes a predictive classification model which performs Weighted Average Ensemble on simple classifiers as well as classifier ensembles such as Logistic Regression (LR), Random Forest (RF), k-nearest neighbours (KNN), Adaboost, Bagging. The proposed model improved the performance up to 99 per cent whereas important base classifier ensembles like RF and Bagging achieved 98 per cent metrics. Other Classifiers like Adaboost and LR attained 97 per cent metrics while KNN went up to 95 per cent. Performance is monitored using various metrics like accuracy, precision and f-1 score.