Busy-hour Traffic Prediction Based On Combined Forecasting Model Of Wavelet Transform

With the rapid development of communication industry,user scale and traffic increase faster, bringing great burden to the communication equipment of mobile operators. Especially in the major festivals, the network paralysis occurs frequently when traffic exceeds the limits of network capacity. Higher prediction accuracy can provide the basis of reasonable network planning and construction for operators. What’s more, it avoids waste caused by excessive investment of operator’s network construction and dissatisfied of customers caused by insufficient investment. Therefore, the busy-hour traffic prediction is one of the important issues and is concerned by many researchers.In the process of traffic prediction, there are several common traffic forecasting methods, such as ARMA model, SVM model, neural network model, Markov model and so on. Although the traditional forecasting models have the advantages, but there are many drawbacks. They only take historical traffic data into account, so it is difficult to accurately predict the complicated traffic. These traditional forecasting models are not suitable for the current traffic prediction. At present, people tend to use the combined forecasting model, it uses the advantages of the single forecasting model to achieve the best results. Two combined forecasting models are proposed in this paper. The research contents are as follows:1. A combined forecasting model considering the influence of multiple factors is proposed in the paper based on wavelet transform、ARMA and PSO-LSSVM. First, four key factors which influence the busy-hour traffic are obtained by correlation analysis on busy-hour traffic data using SPSS. Then wavelet transform is used to decompose and reconstruct the traffic data. The low-frequency component is loaded into ARMA model to predict, while the high-frequency component and the obtained key factors are loaded into PSO-LSSVM model to predict. Finally the forecasting result is achieved by the superposition of predictive values.2. Since the ARIMA model introduces differential method to deal with non-stationary data and in order to improve the prediction accuracy of traffic, the ARIMA model instead of ARMA model. A combined forecasting model based on wavelet transform, ARIMA and PSO-LSSVM is proposed in the paper. Correlation analysis is applied to the busy-hour traffic data using gray correlation method to obtain key factors which influence the busy-hour traffic. By comparing several models, the combined model has higher prediction accuracy and stronger generalization ability.