Network Traffic Prediction Method Of Ensemble Learning Based On Multi-layer Perceptron

With the rapid development of the Internet,users’ demand for network resources is growing.How operators allocate and efficiently use network resources has aroused extensive attention of researchers on traffic prediction.It is the core technology of network traffic prediction in the era of big data to effectively learn and analyze the growth distribution mode of network traffic from a large number of historical network traffic data.It is an effective data processing and prediction method to select the characteristics that meet the actual requirements of different operators according to the characteristics of different network traffic data.In recent years,with the rapid development of deep learning,some deep learning methods show excellent performance in the field of network traffic prediction.In this paper,a spatiotemporal feature prediction method based on multi-layer perceptron ensemble learning is proposed and applied to a city network traffic prediction project in Hubei Province.The main contents include:(1)Convolution neural network is analyzed,and one-dimensional convolution network is used to extract features of network traffic data.Then select a special convolution neural network to extract more spatial information,and finally use the spatial features to establish a spatial model for verification.The experimental results show that the network traffic data has spatial dependence.(2)This paper analyzes the time dependence of network traffic,uses the recurrent neural network to extract the time characteristics of data,and studies some recurrent neural networks with memory function.Finally,the network traffic data is modeled by the gated recurrent neural network(GRU),and the attention mechanism is added to improve the accuracy of the model.(3)After studying ensemble learning,an ensemble learning method based on multilayer perception is proposed.New spatiotemporal features are obtained by integrating spatial features with temporal features.At the same time,in order to speed up the accuracy and convergence speed of the model,teacher forcing mechanism and attention mechanism are added.At last,the traditional deep learning model of LSTM and GRU is compared.The experimental results show that MECG not only has obvious advantages in accuracy,but also has certain advantages in time training cost.Finally,the prediction model based on temporal and spatial characteristics achieves the expected prediction accuracy.The predicted future network traffic data can have a significant impact on the network topology,network information security,network traffic load balancing and other fields.Therefore,network traffic prediction has high research value and broad application scenarios.