Network traffic resource situation predication method based on long short-term memory LSTM model

Abstract

The invention belongs to the technical field of network traffic resource situation prediction and particularly relates to a network traffic resource situation predication method based on a long short-term memory LSTM model. The invention aims to provide a Long Short-Term Memory LSTM cyclic neural network predictive model suitable for a traffic space-time nonlinear characteristic. The model combines the characteristics of strong burstiness and long-range dependence of actual traffic data for model training and quantitative prediction. The method in the invention has the following beneficial effects that the network traffic predication function is realized through the LSTM model, thereby effectively improving the prediction accuracy; and the advantages of the LSTM predication model are moreobvious when the difference of the data samples is large.

Description

technical field [0001] The invention belongs to the technical field of network traffic resource situation prediction, in particular to a network traffic resource situation prediction method based on an LSTM model. Background technique [0002] With the popularization and development of the network, network management faces many challenges. The complex heterogeneous network devices such as wireless sensor network, Ad hoc network, and space-based network make the topology structure increasingly complex. Frequent information interaction makes the network traffic surge, and the complexity and uncertainty description ability of the network operation status is correspondingly reduced. Traffic forecasting has become a core issue in the research of traffic engineering, congestion control and network operation and maintenance management. [0003] Due to the multi-protocol characteristics of the communication network and the burstiness of business sources, the network load often main...

AI Technical Summary

Problems solved by technology

[0005] Communication network traffic will show different characteristics due to factors such as network environment, traffic collection time length, and time scale. It is difficult to find a general modeling method for network traffic prediction. Strong randomness does not appear in strict accordance with certain established rules, and prediction errors are difficult to effectively control through mathematical statistical models; 2) There are many emergencies in network traffic, resulting in a sharp increase in non-linear traffic, even with artificial neural networks Non-linear models such as networks are also difficult to accurately model their changing laws

Images