Hybrid Machine Learning Based Shill Bidding Fraud Detection System Using Recursive Feature Elimination Method

Due to the significant surge in internet traffic and the increasing complexity of networks, manual monitoring of traffic flows and intrusion detection in large networks has become an overwhelming task for network analysts. Maintaining information security is of paramount importance, with fraud detection playing a crucial role in this endeavour. Recently, the adoption of Machine Learning (ML) methods for intrusion detection has also gained substantial popularity. The vast volume of data generated by network traffic often contains irrelevant features, resulting in slower detection processes and diminished system performance. To address this issue, feature elimination or selection is carried out before classification. In this research, a Shill Bidding Fraud Detection system leverages a Hybrid machine learning approach and incorporates a recursive feature elimination method. To ensure a thorough and effective analysis of network traffic for intrusion detection, we combined Support Vector Machine (SVM) and Decision Tree (DT) machine learning models into a hybrid framework. Additionally, we established a multi-class classifier capable of not only distinguishing between malicious and benign traffic but also extending labels to categorize various forms of malicious data. To optimize system performance, This study has identified and eliminated irrelevant features using the Recursive Feature Elimination method. Experimental results demonstrate a significant enhancement in the performance of network intrusion detection.