Network traffic prediction method and device based on data reconstruction and hybrid prediction

Abstract

The invention discloses a network traffic prediction method and device based on data reconstruction and hybrid prediction. The method comprises the following steps: acquiring network traffic value data; reconstructing the network traffic value data to obtain reconstructed network traffic value data; performing decomposition processing on the reconstructed network flow value data by using an EMD algorithm to obtain a plurality of network flow value components; training the GRU-VTD neural network by using the plurality of network flow value components to obtain a trained GRU-VTD neural network; and performing prediction by using the trained GRU-VTD neural network, and calculating a prediction error according to an obtained prediction value and network flow value data so as to perform performance evaluation on the model. The network traffic prediction method based on data reconstruction and hybrid prediction provided by the invention has higher prediction precision and stability.

Description

technical field [0001] The invention belongs to the technical field of network communication, and in particular relates to a network flow prediction method and device based on data reconstruction and hybrid prediction. Background technique [0002] In recent years, people's demand for network services such as instant messaging, search engines, social entertainment, remote office, online transactions, and public services has increased, leading to explosive growth in the scale of network services, and technological progress and user needs have led to more diversified network types. However, due to limited network resources, the continuous increase in network demand will inevitably lead to network congestion and lower service quality. Therefore, it is necessary to grasp the behavior and state of the network to enhance the effectiveness and timeliness of network management. Network traffic is the basis for monitoring network behavior status and researching network behavior. The...

AI Technical Summary

Problems solved by technology

However, these methods rely on prior harmonic basis functions or wavelet basis functions, which are not suitable for dealing with non-stationary nonlinear data

At the same time, due to the large amount of mutation data in the network traffic, the scale of the prediction model is larger, which not only increases the complexity of the prediction, but also reduces the prediction accuracy.

In addition, most of the existing prediction algorithms require real and effective data sets as input, but due to the complexity of network topology, resource constraints of network devices, and high overhead of monitoring high-speed networks, it is impractical to collect all real traffic data in real networks. Practically, this also affects the robustness of the predictive model and thus the stability of the system

Images