Answers are grounded in this paper and cite their sources.
The threat of credit card fraud to financial institutions and their customers is enormous which makes it essential to improve the fraud detection methods.The detection of credit card fraud is a problem for traditional detection methods.This paper presents a new loss function named Regularized Binary Cross Entropy (RBCE) in conjunction with an Autoencoder model.The RBCE loss function aims to improve the traditional BCE loss by incorporating regularization, enhancing the autoencoder's capability to learn robust and meaningful feature representations.This is particularly useful in situations involving high-dimensional and noisy data.The regularization term helps to reduce overfitting by penalizing model complexity, promoting the learning of more generalizable features.Consequently, the autoencoder's ability to identify meaningful features and detect anomalies becomes more accurate.RBCE improves the model's sensitivity to subtle deviations from normal patterns, leading to more precise detection of fraudulent transactions and other anomalies.Two datasets are used in the experiments: European and simulated credit cards.The results show the efficiency of the proposed model in improving fraud detection in both datasets and capturing complex patterns and anomalies in high-dimensional.Moreover, the proposed model outperforms prior works in terms of accuracy (95.130%), detection rate (91.176%), and area under the curve (93.156%) which produces significant improvement in fraud detection.
No takes yet. Share an insight, caveat, or question.
A 2024 study studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: