Network flow multi-step prediction method based on VMD and LSTM

Abstract

The invention discloses a network flow multi-step prediction method based on VMD and LSTM. The method comprises the following steps: generating a training sample; processing the training sample by using VMD to obtain eight modal components; carrying out multi-step prediction model modeling on the modal components by using LSTM, and carrying out network structure optimization by using BN to obtainprediction models of eight modal components and eight new sequences containing features at the next moment; carrying out variational mode decomposition reduction processing on the eight new sequencescontaining the characteristics of the next few moments to obtain a network flow time sequence prediction result containing the next few moments; judging whether the multi-step prediction result of thenetwork flow time sequence meets the accuracy requirement or not; if the LSTM prediction model does not meet the requirements, obtaining a new training sample, performing supplementary training on the LSTM prediction model by using the new training sample, and updating the LSTM prediction model. According to the invention, the defects of accuracy and time performance of a current network flow multi-step prediction model can be improved.

Description

technical field [0001] The invention relates to the technical field of network traffic forecasting, in particular to a network traffic multi-step forecasting method based on variational mode decomposition (VMD) and long-short-term memory network (LSTM). Background technique [0002] Measurements based on time exist in most scientific fields, and data organized by observations are called time series data. The purpose of mining time series data is to find out the statistical characteristics and development regularity of the time series in the sample, construct a time series model, and perform out-of-sample forecasting. Time series data is a data column recorded in chronological order by a unified indicator. All data in the same data column must be of the same caliber and require comparability. Time series data can be period numbers or time points. [0003] Among nonlinear forecasting methods, Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS) and ...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the problem that most existing algorithms in the prior art are based on single-step prediction results combined with multi-step prediction strategies, errors in prediction results will be accumulated during the prediction process, resulting in a significant drop in the accuracy of multi-step predictions. The present invention Provide a multi-step network traffic forecasting method based on variational mode decomposition (VMD) and long-term short-term memory network (LSTM), in order to achieve the accuracy of network traffic multi-step forecasting, and to predict the sequence prediction value of the next few time points , which provides a better and feasible solution for predicting and capturing key information of time series in advance; it improves the accuracy of multi-step forecasting and improves the time efficiency of training

Images