Network traffic forecasting device and method based on neural network

Abstract

The present invention provides a neural network-based network traffic prediction device and method, the device comprising: a source layer processing unit, including a first board-level module, providing source layer nodes, and generating network traffic training data according to the first part of network traffic data; The middle layer processing unit includes a plurality of second board-level modules to provide middle layer nodes, and multiple middle layer nodes perform neural network learning calculations in parallel according to network traffic training data and their respective middle layer weights to generate multiple middle layers The learning result; the target layer processing unit, including the third board-level module, provides the target layer node, and calculates the target layer error item according to the learning results of multiple intermediate layers and the target layer weight; the network traffic prediction unit judges the target layer error Whether the item is within the error threshold range, if so, perform network traffic prediction; source layer nodes, intermediate layer nodes and destination layer nodes communicate according to the network-on-chip protocol. The invention can improve the network traffic prediction efficiency.

Description

technical field [0001] The invention relates to the technical field of the Internet, in particular to a network traffic prediction device and method based on a neural network. Background technique [0002] As the scale of the Internet continues to expand, there are more and more types of network traffic data and services, and the contradiction between supply and demand of network resources and network demand is becoming increasingly acute. Network traffic forecasting is helpful to analyze network security status, manage network scientifically and prevent improper network behavior. Therefore, the research and implementation of network traffic forecasting is of great significance. [0003] The network traffic prediction method based on artificial intelligence neural network has nonlinear and adaptive characteristics, and has high prediction accuracy. Network traffic prediction based on neural network is mainly divided into software implementation and hardware implementation. ...

AI Technical Summary

Problems solved by technology

However, since the dedicated neural network traffic monitoring and forecasting system is implemented by an application-specific integrated circuit, the performance of the algorithm is fixed, and it cannot be adaptively adjusted for the scale of network traffic and different algorithms, and the scalability is poor.

The dedicated neural network traffic prediction platform using multi-core processors adopts the bus control mode of serial access and point-to-point communication, and the bus address resources are limited. Network parallel prediction, while not scalable to network size increase

Images