Impact of Subscription Fraud in Mobile Telecommunication Companies

Abstract

Subscription Fraud (SF) is one of the hardest and most expensive revenue leakage to prevent. This fraud is the leading revenue leakage in telecommunication. The tough economic challenges and saturated telecommunication market in South Africa makes it difficult for telecommunication companies to invest in good fraud prevention systems since their main focus is to increase earnings before interest tax and amortization (EBITA). Most fraud analysts determine the revenue impact of SF on the companies, but less focus is given to the impact on the customer. In this paper, we determine the impact of SF on company’s revenue, churn and customers. The common use of machine learning techniques in telecommunication fraud detection and prevention has been very successful. We used decision tree model to predict SF using post-paid customers data, the model correctly predicted 98 percent of fraud cases and highlighted high monthly subscription fees and payment types as key attributes to the model for predicting fraud, the model further indicated that 30 percent of revenue loss is caused by SF. Information presented in this paper may be used to develop predicting models and also shows a need to develop a solution to vet the customers before contract can be approved.