A comparison of methods for fault prediction in the broadband networks

This paper compares two short-term prediction models of the expected number of faults in broadband telecommunication networks (BB network). These faults occur due to various causes. In order to improve the functionality, various routine or special maintenance (upgrades, replacement of equipment, add new functions, ...) are often carried out in the BB networks on various network elements that may cause unintended and unexpected degradation of services. On the other side of the access and customer part of the network are susceptible to errors induced by various causes, which in the same manner increases the number of faults in the system. This degradation in some cases is not recognized immediately from the systemic alarms, but later they appeared in the form of random disturbances reported by the users of these services. This study examines the prediction of faults by using two different models, Hidden Markov Model (HMM) and the Kalman filter. The model is made on the basis of one-year monitoring of broadband faults analyzed by the services and the exact time of appearance. Assessment of the accuracy of both models is made by comparing the results obtained by modeling and the actual data.