Online Payment Fraud Detection Using Machine Learning
The rapid growth of digital payments, including UPI, cards, net banking and e-wallets, has been accompanied by a sharp rise in online payment fraud. Rule-based filters adapt slowly and raise many false alarms, while fraud itself is extremely rare, so automated detection is a highly class-imbalanced learning problem. This paper presents a complete machine learning framework for online payment fraud detection covering data preprocessing, imbalance handling, model training, threshold selection and evaluation with metrics suited to rare events. Six classifiers (Logistic Regression, Decision Tree, K-Nearest Neighbors, Random Forest, Gradient Boosting and a Multi-Layer Perceptron) and an unsupervised Isolation Forest are compared on a simulated dataset of 100,000 transactions with 1.25% fraud. At the default threshold, Random Forest gave the best F1-score (0.380), precision (0.507) and ROC-AUC (0.837), with a PR-AUC (0.336) practically equal to that of the neural network (0.332), which achieved the highest recall (0.565) at the price of low precision. Further experiments show that SMOTE and random oversampling raise PR-AUC to 0.434 and 0.457, that the decision threshold changes the precision-recall balance more than the choice of model, and that the unsupervised detector is close to random on this data. The results confirm that ensemble methods are well suited to rare-event fraud detection and that accuracy alone is a misleading measure in this domain.
Authors
- Akshay Ajalsonde, Om Auti, Vedant Ghadge
- M. N. Pimple, K. D. Jamkar, Dr. H. B. Jadhav, Ajay Kapse
Publication Details
- Journal
- International Journal of Advanced Research in Science Communication and Technology
- Published
- 2026-11-20
- DOI
- https://doi.org/10.48175/ijarsct-40887
- Primary Topic
- Imbalanced Data Classification Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00