Periodic flow prediction method based on random forest

Abstract

The invention discloses a periodic traffic prediction method based on a random forest, and mainly solves the problem that traffic prediction cannot be performed on different data streams in the prior art. According to the implementation scheme, the method comprises the steps of collecting traffic of various types of application programs; dividing a time sequence for each collected data stream, and counting time sequence point features; calculating a time sequence period index according to the time sequence point features; constructing a period index determination algorithm according to the time sequence period index; calculating time sequence point position features, and constructing a data prediction algorithm in combination with period indexes; and performing flow prediction on the actual data flow by using a period index determination algorithm and a data prediction algorithm. According to the method, the periodic starting point of the data stream is determined, so that the influence generated by non-periodic data is eliminated, and the prediction accuracy is improved; and meanwhile, the time sequence point features of each data stream are counted, and the time sequence point features of different data streams are input, so that data prediction of different data streams is realized, and the method can be used for bandwidth allocation.

Description

technical field [0001] The invention belongs to the technical field of computer networks, and in particular relates to a periodic traffic forecasting method, which can be used for bandwidth allocation in network management. Background technique [0002] Traffic forecasting is based on the transmitted traffic, and according to some characteristics of the selected transmitted traffic, predicts the traffic transmission situation in the future period of time. Traffic forecasting is usually used to determine the benchmark of network traffic to achieve traffic load balancing, or to make reasonable network planning and bandwidth allocation for the network based on characteristics such as bandwidth and delay. [0003] Random forest is a supervised learning algorithm in machine learning that can be used in both classification and regression applications. The traffic prediction mainly adopts the regression mode of the random forest. By dividing the data into a training set and a data...

AI Technical Summary

Problems solved by technology

Among them, the traffic prediction method based on machine learning algorithm can realize the prediction of traffic in a short period of time, but because the current traffic prediction method based on machine learning algorithm mainly predicts the total traffic, it does not predict the traffic of different types Therefore, it is impossible to accurately allocate the required bandwidth for various applications when performing bandwidth allocation; although the traffic classification algorithm based on machine learning can distinguish different types of application traffic, such as the application publication number CN201910201795, a patent application titled "Network Application Recognition Method Based on Multilayer Neural Network", discloses "Network Application Recognition Method Based on Multilayer Neural Network". In this patent, through the neural network algorithm in the machine learning algorithm, Identifying and classifying different network applications can detect the characteristics of different application traffic, but its shortcoming is: only classify different applications to determine the application type, but the actual traffic data size of different applications cannot be determined, and the application cannot be determined. The actual bandwidth requirements of traffic, so the required bandwidth cannot be allocated to applications in network management

Images