Abstract

Telecommunication frauds have been a problem affecting all operators and customers, and is an important factor in their annual revenue losses. Aside from financial impact, it also constrains new service employment and may contribute to adverse costumer perception. detection and prevention of these frauds were focused by academic and proprietary researchers to make the telecommunication industry viable and customer friendly. This paper employed Counter Propagation Neural Network (CPNN) classifier to detect anomaly in subscribers' call records in telecommunication industry. Eighty users' dataset were gotten from a renowned telecommunication industry. The dataset was re-processed to adapt to the neural network processing terrain, and then classified the data into normal and abnormal in an offline processing. The results of performance metrics of the two neural networks showed that (i) for international calls, CPNN produced averages of fraud catching rate and false alarm rate of 0.46 and 0.91 while BPNN 0.54 and 0.84 respectively, (ii) for national calls, CP produced averages of fraud catching rate and false alarm rate of 0.76 and 0.998 while BP produced 0.64 and 0.998 (for both daytime and night) respectively. CPNN produced 69% precision rate better that BPNN which produced precision of 64%. Also, CPNN produced accuracy of 94.2% better than BPNN accuracy of 81.7%.