The Research On Prediction And Control Of Telecommunication Network Congestion Based On The Belief Network Model

The congestion results in the fall of communicating capacity of network. If it can be controlled effectively, the disadvantaged effect of congestion of network will be controlled. Otherwise the local congestion will lead the paralysis of whole network. Therefore the researches on network congestion and seeking the best handling way have not only the academic significance but also practical and economical value.For solving the intelligent prediction and control issue of network congestion, the AI technology is introduced into the network management system, and the belief network modeling of telecommunication network congestion and corresponding decision network modeling are present. The research in this paper focuses on the following aspects based on the network management engineering of Deyang city, Dachuan city and Yibin city of Sichuan province.* The belief network modeling of telecommunication network congestion The belief network is a mathematical tool based on both the probability and graph theory, which gives us an effective approach to represent uncertain knowledge and do flexible inference. It has been hot research topic in diagnosis, decision theory, AI, data mining and so on. There are many uncertainty in telecommunication network and therefore the general modeling of congestion is very complicated and difficulty. The belief network model of telecommunication network congestion in this paper can effectively represent and handle the congestion issue of telecommunication network. * The inference and learning algorithms The inference algorithm for polytree BN and its application in the model are firstly introduced. Secondly, after studying the famous K2 learning algorithm the improved algorithm (MK2) and its parallel implementing method have been present. What is more, the modified MDL algorithm is given. At last the learning method for conditional probability distribution is investigated.* The congestion computing of TN and simulation In this paper a special stochastic process is studied and applied in telephone switch system. The congestion principle is analyzed from link system, telecommunication network and switcher. The congestion computing methods are obtained. The simulation for switcher is conducted. The congestion probability of switcher can be gotten by inference of belief network model with the statistical data of simulation.* The decision network model of congestion control and its applicationThe decision network model for congestion control is constructed in the combination of expert's knowledge with decision theory, and the decision can be made automatically in short time. When the congestion is predicted, it can be controlled quickly.* The application of BN model of telecommunication network congestion An example is present for demonstrating the power of the model to predict the congestion of Deyang city telecommunication network in Sichuan province.The belief network model and decision model in this paper are total new approach to predict and control the congestion of telecommunication network.