Machine Learning-Driven QoT Prediction for Enhanced Optical Networks in DWDM System

Citation

Sakthivel, Sudha and Belgaum, Mohammad Riyaz and Abu Bakar Sajak, Aznida and Alam, Muhammad Mansoor and Mohd Su'ud, Mazliham (2025) Machine Learning-Driven QoT Prediction for Enhanced Optical Networks in DWDM System. IEEE Access, 13. pp. 80445-80460. ISSN 2169-3536

[img] Text
Machine Learning-Driven QoT Prediction for Enhanced Optical Networks in DWDM System.pdf - Published Version
Restricted to Repository staff only

Download (6MB)

Abstract

This research demonstrated the machine learning (ML) classifiers with regression learning to improve an optical system’s quality of transmission (QoT). In Optical Communication, the data can be communicated from source to destination through the established lightpaths. However, as the signal traverses the optical links and devices, its QoT may deteriorate due to various impairments. The QoT is an essential component that determines the connectivity of an optical network. Therefore, ensuring a QoT guarantee is necessary to establish a successful lightpath. Predicting the QoT before establishing lightpaths can guide the routing and allocation of resources required for the lightpaths. In this research, using ML models an appropriate QoT analytical prediction model is developed computationally. Simulations were conducted at a 10 Gbps data rate per channel for 64-channel DWDM systems. The proposed model significantly improves in detecting fiber nonlinearity, and performance was studied using Q-factor, BER, and noise power. The results indicate that the SVM-based classifier with regression learning performs better than any other classifiers discussed in this research. This study assesses the efficiency of the proposed ML models in predicting the QoT for established lightpaths. Results indicate that all the ML classifiers with Regression models can accurately predict the transmission quality for over 90% of lightpaths. However, the proposed SVM-based classifier with a regression model demonstrates superior generalization, with a nearly perfect QoT prediction rate of around 99% for the established lightpaths. In the network planning stage, residual margins are added to compensate for inaccuracies, which ensures accurate signal reception. The proposed ML model achieved a lightpath residual margin with a 0.7dB error.

Item Type: Article
Uncontrolled Keywords: Optical communication, lightpaths, machine learning
Subjects: Q Science > Q Science (General) > Q300-390 Cybernetics
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101-6720 Telecommunication. Including telegraphy, telephone, radio, radar, television
Divisions: Faculty of Computing and Informatics (FCI)
Depositing User: Ms Rosnani Abd Wahab
Date Deposited: 30 May 2025 02:06
Last Modified: 30 May 2025 02:06
URII: http://shdl.mmu.edu.my/id/eprint/13886

Downloads

Downloads per month over past year

View ItemEdit (login required)