Application of Machine Learning in Financial Fraud Detection and Prevention

Detecting fraudulent activities in the financial sector is a critical challenge that requires robust, adaptive approaches. This paper investigates the application of machine learning (ML) algorithms—Logistic Regression, SVM, KNN, Decision Trees and Random Forests—for credit card fraud detection. Utilizing a highly imbalanced dataset, models were evaluated using precision, recall, and F2 score, prioritizing recall to minimize undetected fraud. Our findings demonstrate that Logistic Regression achieved the highest recall (91%), effectively identifying the majority of fraudulent transactions while maintaining a low false-negative rate. SVMs achieved balanced performance with 89% recall, while Random Forests showed superior precision (98%), minimizing false alarms. These results highlight the strengths and trade-offs of ML algorithms for uncovering complex patterns in large-scale financial data and for reducing fraud risk when integrated with real-time detection systems. This research underscores the importance of continuous model optimization using updated data and advanced techniques to counter evolving fraud tactics. By bridging technological innovation with proactive fraud prevention, this paper provides actionable insights for financial institutions, contributing to the development of secure and resilient financial ecosystems.