Network traffic prediction method fusing traffic characteristics

Abstract

The invention discloses a network traffic prediction method fusing traffic characteristics, and particularly relates to the technical field of network information engineering, and the method comprises the steps: collecting and obtaining historical network traffic data in a preset time period range, dividing the obtained historical network traffic data into a preset number of sub-traffic sequences according to a preset time step length, constructing and obtaining a network traffic prediction model for each traffic sequence, and predicting the network traffic corresponding to the sub-traffic sequences by using the network traffic prediction model to obtain a prediction classification label result of the network traffic. Through the technical scheme of the invention, the self-similarity characteristic of the network traffic is used as priori knowledge and is integrated into a gating mechanism of the long-short memory neural network, and the time characteristic of the traffic sequence is extracted by combining the one-dimensional convolutional neural network and the attention mechanism, so that the characteristic of original data can be recovered, and the model prediction result is endowed with interpretability; therefore, the prediction precision of the network traffic is improved, and the change trend of the network traffic is described better.

Description

technical field [0001] The invention relates to the technical field of network information engineering, in particular to a method for predicting network traffic by integrating traffic characteristics. Background technique [0002] In an intelligent network system, accurate and effective prediction can understand the characteristics and changing trends of network traffic in advance, thereby improving the utilization rate of network resources and preventing network congestion. Therefore, it is particularly important to establish an efficient and reliable prediction model for network traffic. The essence of traffic forecasting is to predict time series, that is, to establish a function of its characteristics with respect to time changes based on the historical data of the node to be predicted. Common network traffic forecasting models can be divided into two categories: linear forecasting and nonlinear forecasting. Traditional linear forecasting models include historical aver...

AI Technical Summary

Problems solved by technology

Although the relevant models of neural networks have good predictive performance, deep learning models are usually used as "black box" models. Compared with traditional statistical models, the learning process of deep learning algorithms is opaque and less interpretable.

Images