Performance Prediction in OBSS WLANs Using Machine Learning Approaches

In high-dense deployment of WLANs with very high throughputs in environments like malls, stadiums, colleges, etc., the throughput achieved by next-generation IEEE 802.11 WLANs is much lower than expected. The estimation of throughput through simulators is cumbersome and needs elaborate information regarding the deployment details relating to overlapping basic service sets (OBSS). The objective of this paper to propose well-suited machine learning (ML) models to predict throughput such as artificial neural networks (ANN), k-Nearest Neighbours (KNN) regression, random forest regression, and graph neural networks (GNN). The results obtained in this work is promising to further pursue prediction models based on machine learning. This paper also highlights the use of graph neural networks as a promising approach to solving the elusive problem of performance prediction prior to actual deployment.