Design of Transmission Quality Prediction Procedures on Coarse Wavelength Division Multiplexing with 4-Channel Transmission Using Machine Learning

Eria T. Utamy, Yus Natali, Catur Apriono

Research output: Contribution to journalConference articlepeer-review

Abstract

An optical network is a high-capacity telecommunication network using optical technology and components. CWDM is usually used in optical networks in urban areas because CWDM has a wide bandwidth and follows the needs of urban areas that only need short distances. Machine learning (ML) is a branch of artificial intelligence that is very suitable for dealing with complex problems that are difficult to answer in a reasonable time. Accurate Quality of Transmission (QoT) prediction before the connection establishment is critical for service provision and network resource utilization. The Coarse Wavelength Division Multiplexing (CWDM) model used by the network is by the ITU-T G.694.2 standard, namely a splitting of 20 nm at the wavelengths listed in the standard 1551 nm, 1571 nm, 1591 nm, and 1611 nm. The approach used is a linear regression algorithm with an accuracy of 82.47%, k-nearest neighbor regression with an accuracy of 77.18%, support vector regression with an accuracy of 83.88%, random forest regression of 91.44%, and ANN regression with an accuracy of 94.52%. The random forest regressor is the best machine learning algorithm for predicting transmission quality because even though it has lower accuracy, its computation time far exceeds ANN regression, which is only 1.9098 ms.

Original languageEnglish
Pages (from-to)1-6
Number of pages6
JournalIET Conference Proceedings
Volume2023
Issue number11
DOIs
Publication statusPublished - 2023
Event2023 International Conference on Green Energy, Computing and Intelligent Technology, GEn-CITy 2023 - Hybrid, Iskandar Puteri, Malaysia
Duration: 10 Jul 202312 Jul 2023

Keywords

  • machine learning
  • Optik
  • QoT

Fingerprint

Dive into the research topics of 'Design of Transmission Quality Prediction Procedures on Coarse Wavelength Division Multiplexing with 4-Channel Transmission Using Machine Learning'. Together they form a unique fingerprint.

Cite this