Network Traffic Time Series Forecasting Method Based on Distributed Clustering

Abstract

The invention discloses a network traffic time series prediction method based on distributed clustering. The clustering-based time series prediction method of this method combines the distributed clustering algorithm and the autoregressive model, obtains the time slice tuples by slicing the time series data, and uses the distributed K-average algorithm (k ‑means) clustering algorithm for distributed clustering, and normal distribution fitting for each cluster in the clustering results to obtain a normal distribution N(μ,σ 2 ), μ is the preliminary predicted value. In the prediction stage, the time series data to be predicted is preprocessed to obtain the time slice prefix tuple to be predicted, and the preliminary prediction value corresponding to the nearest cluster is distributedly calculated, and the preliminary prediction value is calculated in combination with the autoregressive model. correction to obtain a more accurate final forecast value.

Description

technical field [0001] The invention relates to the technical field of network flow monitoring, in particular to a method for predicting network flow time series based on distributed clustering. Background technique [0002] In the field of network traffic monitoring, it is very important to analyze and predict network traffic time series data. Currently existing solutions include differential moving average autoregressive model (Autoregressive Integrated MovingAverage model), multilayer perceptron (Multilayer Perceptron), time recurrent neural network (Recurrent Neural Network), long-short term memory artificial neural network (Long-Short Term ), techniques such as clustering-based time series prediction have been extensively studied as possible solutions. From the comparison and measurement of cost, accuracy, energy consumption and scalability, we can analyze that the time series prediction method based on clustering has better advantages in the online environment, and th...

AI Technical Summary

Problems solved by technology

These mainstream technologies work well when the time-series data is flat and the rise and fall are stable. When analyzing and predicting network traffic time-series data with large fluctuations and strong randomness, the prediction effect and prediction accuracy will continue to decline.

At the same time, in actual use, techniques such as temporal recurrent neural networks and long-short-term memory artificial neural networks are computationally intensive and time-consuming, and are not suitable for real-time environments.

Images