Credit Card Fraud Identification with ANN and LSTM based models

The past few years have seen a sharp escalation in the most concerning and widespread type of credit card theft, which has resulted in significant losses. [1] At the same time, security and authentication technology developments are encouraging because fraudsters are using more advanced methods, including card cloning, phishing, skimming, and data breaches. Among the above, many models Common methods such as LR, SVMs, Forest-based algorithms are also utilized, alongside LightGBM are applied for the purpose of fraud prediction. However, due to their respective deficiencies in handling huge datasets, abstracting hierarchical representations, and engineering new features, most of the above-mentioned models could not fare any better than the accuracy level of deep learning models like ANN. It is well known that ANN-based models require a significant amount of training over larger datasets in order to provide the best results, even if they have the added benefit of improving accuracy with improved performance. Since deep learning methods improve detection while adapting to evolving fraud patterns, the work incorporates them into the Credit card scam detection system. Within the end, The research presented here findings demonstrate that the fight against credit card fraud will need to include ongoing advancements in algorithms and data gathering techniques. The obtained results reveal that the classifiers perform more effective in the processed data than in the original data. From all the algorithms, the ANN-LSTM model presented the best results, attaining an accuracy, precision, recall, specificity and f1-score having values of 98%, 97.69%, 97.75%, 97.69% and 97.72% respectively.