Data mining application for fraud behaviour detection in telecoms

Abstract

The main aim of this work was to analyse data, which is gathered in telecommunication, using modern Data Mining techniques. This kind of research activities is very significant for telecoms operators. Annual losses of mobile operators are estimated in millions euro. Lithuanian providers do not publish such statistics. Unhappily for today we do not ave real data, but we hope that this problem is importantt for our providers as well. That's why we used only demonstration data for thi experiment. The solution desribed here permits the early detection of fraud carried out by means of premium telephone numbers. The clustering, also known as discovery Data Mining technique, was chosen to identify customers' behaviour. There are two kinds of tecniques demographic clustering and neural clustering that solve the same problems from different views; both can be used to complement each other. The important elements and results of Data Mining were presented. In this work we used IBM software Inelligent Miner for Data.