Verfahren zur Erkennung von kundenbezogenen Mißbräuchen von Dienstleistungen des Netzbetreibers mittels Online-Analyse von kundenbezogenen Datensätzen

摘要

According to the invention, data sets of a network operator relating to the user behavior in the mobile radio network of honest customers and those with fraudulent intentions are examined on the basis of stochastic models. Beyond the one-dimensional filters that are implemented in MEGS, the invention provides a method for a computer-aided detection of frauds. A principal component analysis provides a graphic tool for easily identifying the data of defrauders by two-dimensional plots and corresponding staining, namely by visual inspection or an automated, computer-based test. According to the invention, an automated test on the basis of Fisher's discriminatory analysis is carried out which identifies simultaneously from among the highly dimensional data sets those data sets that belong to defrauders. Both methods are characterized in that they have small error probabilities of the 1<st> and 2<nd> order while requiring few calculation operations. The invention provides a high detection probability at a low false alarm rate.