Automated troubleshooting of mobile networks based on alarm and performance data using bayesian networks

In this paper, we present an automated troubleshooting method for mobile networks based on alarm and performance data using two Bayesian Networks. The method provides online, reactive diagnosis of malfunction alarms using the Bayesian Network with alarm correlation and diagnosis of cells’ bad performance using another Bayesian network. The proposed method and models were used in a trial with history data of commercial GSM/GPRS network. The results of the trial show that diagnosis accuracy of proposed method is better by about 10% than in current researches, especially in identification of base stations hardware faults - better by about 20%.