REAL TIME BANK TRANSACTION FRAUD DETECTION USING KAFKA AND ML

Abstract

With the rapid growth of digital banking and online transactions, financial fraud has become a major concern for banks and customers. Traditional fraud detection systems often rely on batch processing and are unable to detect fraudulent activities in real time, leading to significant financial losses. This project proposes a real-time bank transaction fraud detection system using Apache Kafka and machine learning techniques to address this challenge. The system utilizes Apache Kafka as a distributed streaming platform where transaction data is continuously generated by a producer and consumed by a consumer. The producer streams transaction data to Kafka topics, while the consumer retrieves the data in real time and passes it to a machine learning model for classification. The dataset undergoes preprocessing steps such as converting nonnumeric data into numeric format, normalization, and feature extraction. A Random Forest algorithm is used for fraud detection due to its robustness and ability to handle complex data patterns. The trained model predicts whether a transaction is normal or fraudulent based on input features. The system is integrated with a web interface using Flask, allowing users to generate transaction streams and view prediction results. Experimental results show that the system can successfully detect fraudulent transactions in real time, although accuracy depends on dataset size and quality. This approach provides a scalable and efficient solution for real-time fraud detection, enhancing security in banking systems and reducing financial risks.