A neural network modal decomposition mechanism in predicting network traffic

Shi Jinmei (2023) A neural network modal decomposition mechanism in predicting network traffic. Doctoral thesis, Universiti Malaysia Sabah.

	Text 24 PAGES.pdf Download (864kB)
	Text FULLTEXT..pdf Restricted to Registered users only Download (6MB)

Abstract

Network traffic prediction is essential for effective network management as it can provide an early warning to the administrator before an incident occurs. This study designs a novel network traffic prediction model namely SAVE-AS. It embeds a new proposed Scalable Artificial Bee Colony (SABC) algorithm, Phase Space Reconstruction, Variational Mode Decomposition (VMD) and an integrated Extreme Learning Machine (ELM). The proposed mechanism starts by using SABC to update the model with a new solution and fine-tune the disturbances in each iteration to deal with the interference in order to find the best values that are also synchronously optimal. The SAVE-AS then constructs an adaptive selection operator. It adaptively selects the number of datasets after VMD optimization decomposition to precisely set the number of hidden layer nodes in an ELM to improve prediction accuracy. Meanwhile, the ELM model is trained using a variety of sub-data sequences that meet the requirements for minimizing computational complexity in modeling. Furthermore, the mechanism eliminates the poor sub-sequence caused by the volatility of the results to accelerate the convergence rate stability. The effectiveness of the model is evaluated using three datasets, i.e. Mackey-Glass, Lorenz chaotic time series of recognized benchmarks and a WIDE backbone of actual network traffic datasets. By comparing six existing model algorithms in all datasets, the results show that SAVE-AS can achieve faster convergence and high predictive accuracy while maintaining stability. Specifically, the predictive accuracy indexes such as Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) can reach a lowest optimum value of 1.1410, 0.1758 and 0.2263, and the average training time is reduced by 25.25%, 23.87% and 41.36%, respectively. The findings demonstrate that the proposed mechanism can predict network traffic more stably, accurately and rapidly in a short time regardless of time intervals or data sequence behavior. Consequently, it can provide effective security warning guidance for network management as well as further improve network service quality.

Item Type:	Thesis (Doctoral)
Keyword:	Network traffic prediction, SAVE-AS, Scalable Artificial Bee Colony, SABC
Subjects:	Q Science > QA Mathematics > QA1-939 Mathematics
Department:	FACULTY > Faculty of Computing and Informatics
Depositing User:	DG MASNIAH AHMAD -
Date Deposited:	10 Jul 2024 11:01
Last Modified:	10 Jul 2024 11:01
URI:	https://eprints.ums.edu.my/id/eprint/39055

Actions (login required)

View Item