Fraud Detection in Bank Transactions Using Machine Learning: A Comparative Analysis of Classification Algorithms

Fraud detection in financial transactions is a critical priority for banks due to the increasing threat posed by fraudulent activities. This project investigates various machine learning models to detect fraudulent transactions in bank payments. The dataset used contains over 59,000 records of both fraudulent and nonfraudulent transactions. The study applies several classification algorithms, including Logistic Regression, K-Nearest Neighbors (KNN), Decision Tree, and Random Forest, to identify the most effective method for fraud detection. The models are evaluated based on accuracy, with Random Forest achieving the highest accuracy of $\mathbf{9 6 . 6 4 \%}$. Techniques such as undersampling are employed to address class imbalances in the dataset. The results demonstrate the potential of machine learning in automating fraud detection and highlight future improvements through more extensive datasets and advanced algorithms.