Online Payment Fraud Detection Using Machine Learning

Abstract

The touchy development of e-commerce has caused an outstanding expansion in computerized misrepresentation, hence jeopardizing monetary strength. In spite of the fact that they are vital, robust anti-fraud systems are now and again hampered by deficient genuine information. We utilized ML models — “Logistic Regression, Decision Tree, Random Forest, Naive Bayes, SVM, ANN, KNN, and boosting techniques like CATBoost, AdaBoost, Gradient Boosting, and XGBoost”— utilizing the  E-Commerce online payment dataset. To further develop discovery, deep learning strategies — including CNNs and a crossover CNN+LSTM model — were likewise used to gather fleeting and spatial examples. Oversampling strategies including SMote were applied to settle information uneven characters. Especially a Voting Classifier integrating Bagging, Random Forest, and Boosted Decision Tree, gathering approaches accomplished the best accuracy of 97%. The CNN+LSTM model better fraud pattern recognition even more. The innovation quickly messages alert after seeing false movement in web-based installments, subsequently working with convenient mediation for additional security. This paper shows how refined machine learning and deep learning strategies could uphold fraud detection in the quick growing e-commerce area.