Abstract
The recent focus on monitoring and managing telecommunication networks in a more efficient and autonomic way has led to the widespread application of machine learning (ML) approaches for network management tasks. In order to study the behaviour and evaluate the performance of such network management systems, it is a requirement that a suitable modelling framework exists. The work presented here addresses this need by comparing existing ML-based approaches and proposing a solution which employs the prediction capabilities of the Bayesian networks (BN) approach. It also formulates a BN-based decision support system for providing real-time call admission control (CAC) decisions in the next generation network (NGN) environment. In order to provide a realistic simulation environment, it surveys the existing computer network simulators and BN modelling simulators to choose the most suitable simulators to test the proposed models. The novelty of this research is validated through offline modelling and online performance evaluation of Bayesian networks-based admission control (BNAC) in terms of the metrics of packet delay, packet loss, queue size and blocking probability. This paper concludes that BNAC approach is appropriate choice for implementing a CAC solution which is efficient and autonomic.