Design and Implementation of a Web-Based Credit Card Fraud Detection System

Credit card fraud detection is a key problem in various industries, including banking, e-commerce, and healthcare, where detecting fraudulent activity in a timely manner is crucial. This research project describes a unique method for developing a credit card fraud detection system using a deep learning algorithm, an Autoencoder (AE). The system includes a user-friendly Graphical User Interface (GUI) built with a Streamlit web application framework that allows users to input data and obtain real-time fraud detection. The developed system uses the Autoencoder (AE) architecture, which can learn complicated hierarchical data representations. The model is trained using valid transaction data to capture the underlying patterns and properties of authentic transactions. When the model meets data that deviates considerably from its learned representations, anomalies, indicating probable fraudulent actions, are recognized. Multiple assessment criteria were used to analyze the system's efficacy, including precision, recall, confusion matrix analysis, model reconstruction error, accuracy, and F1 score. The precision assesses the proportion of accurately recognized fraud cases among all anticipated fraud instances, whereas recall assesses the actual fraud cases found. The confusion matrix analysis offers a detailed picture of the model's performance across various prediction outcomes. The developed credit card fraud detection system was created and trained with Google Collab, a cloud-based platform that provided the computing resources needed for the Autoencoder (AE) model training. The combination of the GUI-based user input system and the Autoencoder (AE) model allows users to engage with the fraud detection system naturally, allowing for real-time detection of possible fraud threats. The reconstruction error of the model is an important measure of its capacity to help discriminate between regular and fraudulent cases. The model achieved an accuracy of 0.92, demonstrating the system's ability to predict both non- fraudulent and fraudulent online credit card transactions.