Telecom Fraud Detection with Machine Learning on Imbalanced Dataset

This paper compares the performance of different ML algorithms to a fraud dataset. Telecom operators have long history and experience against fraud activities. During the time and changing the role of telecom operators, from service and infrastructure carriers to communication service providers handling data, voice, and content transfer. There is an urgent need to develop efficient and adaptive algorithms for early fraud detection and prevention. In this paper, we used machine learning techniques as an effective method to detect fraudsters in mobile communications. Fraud datasets are derived from the real telco operator's environment. Different experiments with a few popular machine learning classification algorithms were performed to evaluate the performance of the model. The used dataset, characterized by highly imbalanced distribution, is oversampled with SMOTE method to generate synthetic minority class instances and provide experiments. Performance metrics are evaluated and future research directives are proposed.