Traffic Forecasting Based On VMPSO-BP Neural Network Algorithm

Traffic prediction, it is to point to the first analysis and comprehensive historical traffic characteristics and some specific factors affect traffic, then use a suitable method to predict future traffic. As the mobile communication network conveys the number of traffic and its users into increasingly rising, then to make mobile communication network can have a long-term and stable operation, It needs timely and effectively for mobile communication network to carry on the long-term planning and reasonably optimization. Due to traffic prediction is known to exist with the data to predict future uncertain and random events, so it needs to extract the features and rules known events, using some proper prediction model and related technologies to achieve.In the mobile network traffic prediction process, after practice and suming up the steady accumulation,There has the following several more common prediction methods: time series prediction method, regression analysis forecasting methods and neural network prediction method, etc. But traffic data is complexy and affecting traffic factors show a very strong nonlinear characteristics, so according to all aspects of the characteristics and research of factors which affect it in order to find a suitable model to forecast, and try to minimize the predicting error. This thesis based on BP neural network is proposed a new improved algorithm, and the research content is as follows:1.Based on the original particle swarm algorithm it forms an improved variants of the new algorithm,it is the speed variation of particle swarm algorithm.it is different from the traditional variation PSO algorithm. All particles of the original PSO algorithm tend to the same points,with particle swarm current place again for the center to mass population initialization particles, so that later, in keeping with their respective each particle history of under the premise of search inertia will be in a new position movement, the traditional variation of particle algorithm is the position variation, but the algorithm of particles is the speed of the mutation.2.Then combining the algorithm with the BP neural network, initializationing BP network weights and threshold,with the top of the proposed algorithm weights and threshold are optimized,after optimization the BP network is used to predict mobile traffic.Experimental results show that forecast precision of this method is high,and the effect of prediction is very good.