A Performance Study of a Modified Grey Network Traffic Prediction Mechanism

Abstract

This study focuses on a modified grey theory framework of network traffic prediction. Grey's framework is a predictive tool in many areas, such as network traffic forecasting, weather forecasting, economic impact assessment, and more. Several types of studies have carried out this theory on the prediction of network traffic. While the researcher can also use the neural networks and machine learning to predict network traffic, these methods involve a significant volume of data to forecast network traffic reliably. On the other side, Grey theory Method may use very little information (4 or more data) to evaluate uncertain data. The strength has inspired us to modify the Grey theory system for the prediction of network traffic. This research modified the traditional grey model, GM(1,1) algorithm and tested the mechanism's predictability. The DARPA 1999 Week 1 data set was mounted to a modified grey model, GM(1,1) with z = 1.0. The feature selection is used to remove the irrelevant feature. The findings of the experiment revealed that the performance of the modified grey model, GM(1,1) with z = 1.0 is more effective than the traditional grey model, GM(1,1) with z = 0.5. The prediction accuracy for the traditional grey model, GM(1,1) with z = 0.5 is 92.38% while the modified grey model, GM (1,1) with z = 1.0 is 94.10%. The improvement rate for the modified grey model, GM(1,1) with z = 1.0 is 1.72% compared to the traditional grey model, GM(1,1) with z = 0.5.