Fault Prediction in Self-Healing Telecommunication Systems

Abstract

Telecommunication systems are heterogeneous networks with parts supplied by many vendors. Such complex systems face a number of faults that may deny services to the end users resulting in revenue losses to the telecommunication companies. Best case scenario is to avoid these faults completely, or failing that, correct the faults as soon as possible. Therefore there is a need for self-healing networks that can proactively predict and correct faults automatically. In this paper a fault prediction technique is presented that is useful in a self-healing network. The proposed technique first trains an artificial intelligence technique on the historical alarm data to find correlations and then uses these correlations to predict future alarms. The artificial intelligence techniques being used are, Artificial neural network, support vector machine, Kalman filter and hidden Markov model. In this paper we reported on artificial neural network. The proposed technique is applied on the alarm data from a real telecommunication company and prediction accuracies of the proposed technique are calculated. The details of the proposed fault prediction technique and results that suggest optimal parameters are presented. The proposed technique is effective in a proactive self-healing network.