FRAUD DETECTION IN FINANCIAL TRANSACTIONS
Areej Alsulami, Rana Alabdan · Advances and Applications in Statistics · 2024
Incidents of fraud are increasing intensely considering the growth of modern technology. Globally, the increase has led to the loss of billions of dollars annually. Though, the use of preventive technologies remains the most effective measure to reduce fraud, it is crucial to understand that fraudsters are vastly flexible and are ready to explore methods to escape such measures. Using machine learning, we emphasis on an advanced and innovative fraud detection system. The study uses historical transaction statistics in training predictive models to recognize fraudulent transactions and lessen incorrect positives correctly. This study focuses on aiding the reduction in financial fraud and improving the security and dependability of digital financial transactions by merging data preprocessing, feature engineering, and model evaluation. The paper uses machine learning models such as logistic regression, CatBoost, decision trees (DTs), and random forests to compare performances and pick the best one for the study case. Accuracy, F-scores, precision, recall, and support values determine model performance. The results in this paper indicate that DT and CatBoost provide the highest level of accuracy, at 99.9% and 99%, respectively. These discoveries establish the potential of machine learning in significantly improving fraud detection, thereby curbing financial losses and improving the security of digital transactions. Received: April 8, 2024Accepted: May 14, 2024