Enhancing the Detection of Debit Card Fraud Detection Using Logistic Regression and Random Forest Techniques

Debit card fraud is one of the major financial crimes globally, causing a very great financial losses for financial institutions and individuals. The traditional mode of fraud detection systems often struggles to keep with the latest change in fraud patterns, due to the dynamism of the criminals resulting in high rates of false positives. This project proposes an improved system based on machine learning models to accurately and effectively identify fraudulent transactions. With machine learning models, fraudulent activities can be monitored and identified in real time. It is able to adapt to the changing nature or approach of fraudsters due to advancement in technology unlike the traditional model that is static in nature. Machine learning approach to fraud detection will mitigate the instances of false positive. This project focuses on utilizing machine learning algorithms, namely Random Forest (RF) and Logistic Regression for detecting debit card fraud. A series of rigorous experiments were conducted to evaluate the effectiveness of RF and LR in detecting debit card fraud. Evaluation is carried out using various performance metrics, including accuracy, precision, recall, sensitivity, specificity and F1-Measure.