XAI-Guided Optimization of a Multilayer Network Regression Model

Katarzyna Duszynska, Pawel Polski, Michal Wlosek, Aleksandra Knapinska, Piotr Lechowicz, Krzysztof Walkowiak

Abstract: The recent technological advances create increased network capacity demand, highlighting the need for new network optimization methods. However, the proposed solutions require broad testing with numerous time-consuming simulations. Thus, estimation methods based on Machine Learning (ML) are developed to improve this process. In this work, we create a regression network model to predict four resource utilization metrics using the input set of connection requests. Using eXplainable Artificial Intelligence (XAI) tools, we optimize the proposed model for faster inference without a decrease in prediction quality.

External IDs:dblp:conf/networking/DuszynskaPWKLW24