Ensemble Learning and Deep Neural Networks for Enhanced Credit Card Fraud Detection

Higher digital payment volume and escalating fraud patterns create obstacles for the financial industry when detecting credit card fraud. Machine learning approaches enable the sector to analyse buying patterns, which enables the identification of irregular conduct to establish detection accuracy. The study entity conducts model comparison through a Multi-Layer Perceptron (MLP) Neural Network by using two pairings of classifiers that include XG Boost Classifier with Light GBM Classifier and Random Forest Classifier along with AdaBoost Classifier and Cat Boost Classifier. Previous data processing enables exploratory data analysis (EDA) to pull vital details from the datasets. The success rate of classification depends on two evaluation approaches that combine AUC-ROC scores and confusion matrices for performance assessment. The evaluated results show that the XG Boost and Multi-Layer Perceptron Neural Network method reaches an accuracy rate of 97.96%, hence constituting an optimal system for detecting fraudulent transactions. Digital fraud systems obtain enhanced security through ensemble learning that works along with deep learning frameworks.