Credit Card Fraud Detection using Machine Learning

This project presents a credit card fraud detection system designed to analyze transaction data for identifying fraudulent activities. Using a dataset of European cardholder transactions from September 2013, it examines 284,807 transactions over two days, of which only 492 were labeled as fraudulent, making the dataset highly imbalanced, with fraud cases constituting just 0.172% of all entries. Each transaction is represented by 30 numerical features, including 28 principal components (V1–V28) derived from a PCA transformation for confidentiality. Additional fields include 'Time,' indicating seconds elapsed since the first transaction, and 'Amount,' reflecting the transaction amount, both potentially relevant for cost-sensitive analysis. The system uses a Random Forest algorithm, implemented in MATLAB, for classification. During operation, users input transaction details into the system, which then processes the data to detect anomalies. If fraud is suspected, the system flags the transaction as "Suspicious Activity Detected" and advises contacting the bank. For genuine transactions, the message "Details Verified and Processed" confirms a successful transaction. This approach leverages machine learning to provide a reliable and efficient detection mechanism for real-time fraud prevention.