Research And Implementation Based On Traffic Prediction And Load-Balanceing Of Congnitivennetwork

As future Internet structures and user behavior patterns undertake earth-shaking changes, network traffic volume will also occur a tremendous growth with the accumulated user immenseness and service type diversity. With regard to the increasing complexity of network access environment, the state of the network environment based on how effective load balancing network traffic prediction and decision-making, better coordination of various network resources, to solve network congestion and quality of service degradation and other issues, the future of heterogeneous network traffic control and performance measurement has important significance. Cognitive network, not only owns dynamic spectrum sensing and switching ability, but also possesses cognitive learning and decision-making functions by virtue of all kinds of business and quality of service information, such as node geographic information, link quality, network traffic, user preferences, and other aspects of cognitive learning decision-making skills. Thus, we can take advatages of the strengths of cognitive network to improve the traffic prediction accuracy and load balancing, further improve interworking, optimize network resource and improve wireless network performance.Based on the analysis above, after having studied the traditional Internet traffic prediction technology and cognitive networks features, we focus on to imvestigate the traffic prediction and load balancing technology in cognitive network, carried out this project work, the main contribution is as follows:Firstly, we studied some traffic prediction method of the current network, the ARIMA model and Markov chain model. Then analyzed their principles and use simulation models to predict these experiments. Then we proposed a traffic prediction model in cognitive network based on Partially Observable Markov Process. The method takes into account the unpredictable and incompleteness of the cognitive network, according to the state in last time period, using parameters that can get, solve the significant parameter selection problems, increase traffic prediction accuracy.Secondly, we proposed a load balancing algorithm in cognitive network, where traffic can be balanced in accordance with services classification and traffic prediction, improve network throughput, improve network resource utilization and data transmission performance.Lastly, we use such model in the cognitive network system, and compare it with the traditional mode, selecting the network throughput, packet loss rate, delay, jitter to do analysis, then data shows that the model will lead a better result.