Enhancing credit card fraud detection using machine learning: A comparative analysis of SVM and MLP

Credit card fraud has long been a significant issue globally and within Europe, causing substantial financial losses globally. To mitigate the problem, this study aims to create predictive models that are capable of accurately identifying fraudulent credit card transactions, thereby reducing economic losses. The dataset utilized in this study was sourced from Kaggle and includes 284,807 credit card transactions from Europe recorded in September 2013, with only 0.172% of transactions identified as fraudulent. This study utilized Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP), also known as neural network) to develop two fraud detection models, with K-Fold Cross-Validation employed to evaluate their performance. The SVM model achieved an AUC-ROC score of 72.43%, while the MLP model demonstrated superior performance with an AUC-ROC score of 93.56%. Further sensitivity analysis of the MLP model revealed that with a false negative rate of 0.227, potential financial losses could amount to approximately €60,968.92 over two days of transactions in European countries. These findings highlight the critical importance of deploying highly accurate models for fraud detection, highlighting the MLP model’s efficacy in significantly reducing the financial impact of credit card fraud.