An Automatic ATM Card Fraud Detection Using Advanced Security Model Based on AOA-CNN-XGBoost Approach

The implications of financial fraud are becoming more significant as the risk of it grows. For two main reasons, data mining for the purpose of detecting ATM card fraud is quite difficult. First, datasets about ATM card fraud frequently display significant bias. Second, what constitutes legitimate and fraudulent behavior is defined differently every time. The variables chosen and the sample methods used have a direct bearing on how well ATM card fraud detection algorithms work. Card fraud at ATMs is a major threat to the security of online payment systems and banking. Model training, feature reduction, preprocessing, and procedure selection must all be part of a systematic approach. The importance of preprocessing in avoiding inaccurate model output and keeping data quality high cannot be overstated. As part of the preprocessing step known as feature reduction, correlation-based measures are used to examine feature correlation. After that, the model is trained using AOA-CNN-XGBoost, which produces remarkable outcomes with a 97.13% accuracy rate.