Effective Anomaly Detection of E-Wallet Transactions Using Deep Learning and Spatio-Temporal Analysis

Electronic wallet transactions have become integral to daily life, yet this increase also escalates the risk of anomalies and fraud. This study develops an anomaly detection model for e-wallet transactions using Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) models, enhanced by spatio-temporal analysis. Transaction data from the Mony platform, including transaction IDs, timestamps, amounts, and geographical coordinates, were pre-processed, features were extracted, and data labeled to identify normal and anomalous transactions. The models, trained with SMOTE to address data imbalance and evaluated using accuracy, precision, recall, and F1-score, demonstrated that the LSTM model outperforms others with a test accuracy of 94.64%, precision of 94.81%, recall of 94.64%, and an F1-score of 94.62%. The GRU model followed closely, while the RNN model showed lower performance. The study highlights the effectiveness of spatio-temporal features in anomaly detection and establishes the LSTM model as a robust tool for improving transaction security.