Optimizing credit card fraud detection with RNN and ensemble machine learning models

Credit card fraud remains a major problem facing the world’s financial industry. Owing to advancement in technology and its adoption in financial transactions credit cards has become an essential and important channel for consumers and merchants. Hence, while using of the internet has become widespread at the moment, it has more opportunities relate to fraudsters. To tackle this, mitigated sophisticated approaches involving deep learning and machine learning are required for fraud detection in its early stages and to avoid as much false positives as possible. This research aims at applying the RNN model interlinked with the machine learning algorithms for improvement of the fraud detection. Moreover, the issues with data distribution were addressed using random sampling, and feature engineering as well as ensemble decision making was incorporated to achieve optimum model performance. The dataset was split into training and test datasets where RNNs were employed to extract useful features and achieve classification using other machine learning algorithms namely XGBoost, RF and KNN. Statistical and analytical methods for the model were implemented using Python, resulting in an outstanding overall accuracy of 99.6%; precision of 99.3%; recall measurement of 100%; F1 score of 99.6%; and ROC AUC score of 99.6%. In conclusion The present research work represents the implementation of an efficient RNN-based fraud detection model coupled with machine learning algorithms that provides excellent accuracy of 99.6%, while precision, recall, and F1 scores are equally impressive. Feature engineering, random sampling, and ensemble decision-making employed by the model reduce the false positive rate. This approach is promising in early fraud detection in financial transactions, enhancing security and accuracy.