DBSCAN Clustering for User Pairing in Wireless Networks
Sharan Mourya, Pavan Reddy, SaiDhiraj Amuru, Kiran Kuchi · 2025
In recent years, machine learning techniques have gained prominence as effective solutions for real-time wireless resource allocation problems that are known to be NP-hard. One specific problem in wireless communication systems is user pairing, which involves selecting users to be scheduled together while optimizing interference reduction and throughput maximization. In this study, we conduct a comprehensive analysis and comparison of various user pairing algorithms, identifying their limitations. We propose an unsupervised learning approach for user clustering utilizing the density-based spatial clustering of applications with noise (DBSCAN) algorithm. Our proposed method surpasses other scheduling approaches, including k-means clustering-based user pairing and semi-orthogonal user scheduling (SUS), demonstrating significantly superior performance. We provide a comprehensive evaluation of each method’s performance, stability, complexity, and latency. Notably, at a signal-to-noise ratio (SNR) of 20dB, our proposed method outperformed the conventional k-means algorithm by 33.8% while consuming 42.2% less resources.