Application of Ensemble Regression Methods in Elastic Optical Network Optimization

Poster (konferens), 2024

In this paper, we tackle the problem of estimating the operation of resource allocation algorithms in multilayer optical networks. We show that it is possible to create a regression model simulating a routing and spectrum allocation algorithm to predict four different metrics (i.e., highest occupied slot, average highest occupied slot, sum of occupied slots, and number of active transceivers), only having the input set of connection requests and no information about the underlying topology. We analyze the performance of various ensemble methods, including XGBoost, Random Forest, and stack- ing models on two large topologies, and demonstrate their good prediction capabilities.