Machine learning algorithm estimation and comparison of live network values of the inputs which have the most effect on the FEC parameter in DWDM systems


Creative Commons License

YÜCEL M., Osmanca M. S., Mercimek I. F.

JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI, vol.27, pp.133-138, 2024 (ESCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 27
  • Publication Date: 2024
  • Doi Number: 10.2339/politeknik.1109101
  • Journal Name: JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI
  • Journal Indexes: Emerging Sources Citation Index (ESCI), TR DİZİN (ULAKBİM)
  • Page Numbers: pp.133-138
  • Keywords: DWDM performance, performance measurement, FEC, machine learning, WAVELENGTH
  • Gazi University Affiliated: Yes

Abstract

The purpose of this study is to determine the effect of 7 different algorithms on the FEC value, which is one of the most important parameters of the quality measurement metric in DWDM networks, analyzing these changes through machine learning algorithms has determined which parameter is the most important input affecting the FEC parameter according to the live network values. To determine the algorithm that gives the most accurate FEC value according to the estimation results in machine learning, it is aimed to make analyzes vendor agnostic. As a result; In this analysis, which was conducted with 945 live network values from 3 different vendors, it was determined that the most important parameters affecting the FEC value are the number of channels, fiber attenuation, and fiber distance, and these parameters were estimated most accurately with the decision tree machine learning algorithm.