Credit Card Fraud Detection Using Deep Neural Networks with Autoencoders and SMOTE Techniques

This project aims to detect credit card fraud using deep neural networks (DNN) combined with Autoencoders and Synthetic Minority Over-sampling Technique (SMOTE). Credit card fraud detection is a critical challenge due to the large and imbalanced nature of transactional datasets, where fraudulent transactions are much fewer than legitimate ones. The system leverages a combination of DNN models, including Feedforward Neural Networks (FNN), Convolutional Neural Networks (CNN), Recurrent Neural Network (RNN), XGBoost, and Random Forest classifiers, to classify transactions as either fraudulent or non-fraudulent. Autoencoders are employed to detect outliers or anomalies in the data, which represent fraudulent activities. SMOTE is applied to address class imbalance by generating synthetic samples of the minority class (fraudulent transactions), improving the model's ability to detect fraud. The project is implemented using a Flask web framework, where users can upload transaction data for prediction and view model performance metrics. The system provides a robust and scalable solution for credit card fraud detection, leveraging machine learning and deep learning techniques to achieve high accuracy, precision, and recall. The user-friendly interface allows for easy deployment in real-time fraud detection applications.

Keywords: Credit Card Fraud Detection, Deep Neural Networks, Autoencoders, SMOTE, Anomaly Detection, Flask, XGBoost, Random Forest, CNN, FNN, RNN, Imbalanced Data, Machine Learning, Predictive Analytics.