Comparative Data Oversampling Techniques with Deep Learning Algorithms for Credit Card Fraud Detection

Boulbaba Ben Ammar, Mohamed Ben Aouicha
December 2022
PDF Cite DOI

Abstract

Financial businesses have a significant challenge with credit card fraud. Approximately more than one million individuals from different countries are losing money yearly due to credit card fraud. Due to privacy concerns, there is an absence of research studies that can be conducted on the analysis of real credit card data. In this study, various deep learning algorithms were applied for credit card fraud detection using a real credit card dataset. The methodology presented in this paper has also been focused on comparative of performance testing of different data resampling based on oversampling techniques like Random, SMOTE, BorderLine SMOTE, SVMSMOTE and ADASYN along with deep learning neural networks inculuded LSTM, MLP, and ANN models. For verifying the performance of the proposed models to detect fraud, the used dataset was splitted into 10% validation, 20% testing and 70% training. Furthermore, standard performance metrics such as accuracy, sensitivity, precision, and f1-score metrics were employed to assess and selects the best model. After comparing the results of the experiments performed for fraud detection, it was found that the LSTM model provided the best results for credit card fraud detection with achieving 99.9% accuracy, compared to the other models.

Type
Conference paper
Credit Card Financial Businesses Oversampling Deep Learning
Boulbaba Ben Ammar
Boulbaba Ben Ammar
Assistant Professor

Boulbaba Ben Ammar is an Assistant Professor at the Computer Science Department, University of Sfax, Tunisia.

Mohamed Ben Aouicha
Mohamed Ben Aouicha
Professor

My research interests concern information retrieval, semantic technologies, social media analytics, knowledge representation, Big Data and graph embedding.