Abstract: This article explores an innovative methodology for credit card fraud detection, employing Autoencoder Neural Networks as a powerful tool. This study focuses on enhancing anomaly detection systems. Leveraging TensorFlow and Keras, the Autoencoder model is trained in an unsupervised manner, concentrating exclusively on normal transactions. This approach allows the model to learn the inherent patterns of legitimate transactions, enabling effective identification of potential fraud. The training dataset, encompassing two days of credit card transactions (284,807 instances with 492 labeled as fraudulent), reveals a highly imbalanced distribution. Through meticulous data exploration, insights into transaction amounts and timestamps are gained, informing the subsequent model architecture. The Autoencoder, comprising four fully connected layers with L1 regularization, demonstrates its efficacy in capturing the underlying structure of normal transactions. By evaluating the reconstruction error as a key metric, this project showcases the promising potential of Autoencoder Neural Networks in significantly improving credit card fraud detection mechanisms.
Keywords: CNN Autoencoder Neural Networks, Anomaly Detection, Credit Card Transactions, Fraud Detection, Deep Learning
Cite:
Mr.A Vishnu vardhan, Muppiri P V S N M L Ankitha, Pasupuleti Divya Sri, Mohana Battula, Muppuri Venkata Triveni Sai Priyanka," Anomaly Detection in Credit Card Transactions using Autoencoders ", IJARCCE International Journal of Advanced Research in Computer and Communication Engineering, vol. 13, no. 3, 2024, Crossref https://doi.org/10.17148/IJARCCE.2024.13320.
|
DOI:
10.17148/IJARCCE.2024.13320
