Machine learning has emerged as a powerful tool for both engineering and geo-localization applications. In this study, we investigate the Terabit/sec bandwidth wireless technology application using specialized ns-3 simulation tools. Through extensive simulations, we explore various scenarios with diverse parameters, including population density, topology types, and overlapping ratios among consecutive radio sectors centered around a single access point. To extract meaningful insights from the data, we employ the DBScan unsupervised learning method, enabling us to identify the optimal number of classes for sector efficiency features. Our optimization approach considers both the number of outliers and the minimum number of elements within each radio sector. By analysing a synthetic dataset generated from the simulation cases, we uncover valuable insights and establish the optimal working point for the system.