Machine Learning–Based Traffic Prediction for Peering Selection in ISP Networks
Laakso, Mikko (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025120532846
https://urn.fi/URN:NBN:fi:amk-2025120532846
Tiivistelmä
The work was commissioned by Elisa Oyj, a Finland-based telecommunications company. The thesis aims to improve Elisa’s utilization of private peering vs. public peering or transit providers to reduce networking costs.
The goal for this thesis was to develop a program which uses machine learning to examine past traffic patterns in order to predict future traffic amounts to enable better utilization of investment and proactive building of connectivity.
The research method used was a methods approach that combined quantitative analysis, qualitive insights and applied research.
To understand the topic, the thesis researches the structure of the Internet and how traffic is passed between different parties and the costs associated with these methods. The following research questions are answered: Can we use legacy data to predict future trends? What factors do we need to consider to make good predictions?
The defined research questions were answered in the literature review.
The thesis reaches the set goal of creating a program which utilizes different machine learning models to predict future traffic amounts between Elisa and possible peering partners.
The goal for this thesis was to develop a program which uses machine learning to examine past traffic patterns in order to predict future traffic amounts to enable better utilization of investment and proactive building of connectivity.
The research method used was a methods approach that combined quantitative analysis, qualitive insights and applied research.
To understand the topic, the thesis researches the structure of the Internet and how traffic is passed between different parties and the costs associated with these methods. The following research questions are answered: Can we use legacy data to predict future trends? What factors do we need to consider to make good predictions?
The defined research questions were answered in the literature review.
The thesis reaches the set goal of creating a program which utilizes different machine learning models to predict future traffic amounts between Elisa and possible peering partners.