Unsupervised fraud detection in banking transactions using fuzzy clustering and outlier detection techniques

The increasing prevalence of online banking and e-commerce has led to a significant rise in financial transactions, which has consequently increased the risk of fraudulent activities. Detecting fraud in banking transactions remains a major challenge, especially due to the scarcity of labeled data and the evolving nature of fraudulent behaviors. This paper proposes an unsupervised machine learning approach that combines fuzzy clustering and outlier detection techniques to enhance fraud detection. The proposed method begins with Fuzzy C-Means (FCM) clustering, grouping transactions based on their characteristics while assigning degrees of membership that capture behavioral uncertainty. Transactions with low membership in normal clusters are flagged as potential fraud. Additionally, Isolation Forest and Local Outlier Factor (LOF) are employed to detect rare and anomalous patterns characteristic of fraudulent activities. To evaluate the model, we use a large-scale real-world banking dataset containing ۱,۰۰۰,۰۰۰ transactions, each represented by seven key features including transactional, temporal, and behavioral attributes. Results show that combining fuzzy clustering with outlier detection significantly improves fraud detection accuracy, achieving a high true positive rate (TPR) and a low false positive rate (FPR). The method is unsupervised and suitable for environments where labeled data are scarce or costly, providing a robust and scalable solution for real-time fraud detection.

کلیدواژه ها:

fraud detection ، Unsupervised Learning ، Fuzzy Clustering ، outlier detection ، banking transactions ، Fuzzy C-Means ، Isolation Forest ، Local outlier factor ، Machine Learning ، Anomaly Detection