Intelligent Detection of E-Wallet Transaction Fraud Using Hybrid Deep Learning and Ensemble Machine Learning Models

The fast development of the digital payment systems has created more danger of fraudulent transactions in e-wallet services. In this study, the proposed fraud detection system is an integrated system of machine learning and deep learning to improve the accuracy of fraud detection. The dataset used in developing models was a synthetic mobile money transaction dataset on Kaggle. Preprocessing of the data entailed the elimination of redundant attributes, label encoding and scaling the features with MinMaxScaler and selecting the features with Lasso regression. Several machine learning models were tried such as the Random Forest, K-Nearest Neighbors, Decision Tree, Logistic Regression, and XGBoost. An ensemble model was also implemented between RF and XGB. Moreover, the deep learning schemes such as Artificial Neural Network, Convolutional Neural Network, Recurrent Neural Network and a hybrid CNN+RNN model were created and optimized through hyperparameters. The results demonstrate superior performance from the CNN (92.24%) and hybrid CNN+RNN (92.22%) models, outperforming traditional machine learning approaches. This hybrid deep learning system offers a powerful and expandable solution to real-time fraud detection in mobile payment systems, which enhances safe and reliable digital financial systems.