Identifying Fraudulent Activity in Machine Learning for Telecommunications Employing in CDR

Abstract

Telecommunication scams are a global issue that costs many customers and service provider businesses a significant amount each year. We provide an efficient and suitable scam user identification technique that depends on the customer's Call Detail Record (CDR) for quick and affordable identification of telecommunication scam clients. The techniques for classification employed in network learning as well as in several scientific and technical domains, such as computer vision, understanding speech, email virus detection, etc., served as the inspiration for the studies we did. The ML and pattern recognition modules make up the two halves of the suggested approach. The ML module uses summary features to identify people using the Support Vector Machine (SVM) technique, which is based on supervised learning. The model identification unit analyzes suspect individuals produced by the machine's instructor using a Finite State Machine (FSM) according to scam user activity. Unauthorized users can be identified after using both components. We implement our approach and evaluate it with actual data. The trials show that the proposed method can achieve exact detection and outperforms the state-of-the-art techniques.