Automatic detection and prediction of signal strength degradation in urban areas using data-driven machine learning

Abstract

Signal strength degradation represents a variation in the coverage area for radio networks. Building maps that represent this degradation requires collecting information about signal coverage in scattered locations, which can be done conventionally by measurement methods such as the manual drive test. Nevertheless, as this process is large-scale, time-consuming, and costly, several methods for the minimization of drive tests have been introduced. In this study, our methodology first consisted of dividing the study area into several zones, and each zone into several sectors without considering the position of the existing broadcast base station. Then, we developed a custom mobile application to collect the signal strength data and the location coordinates of the concerned area. For data collection, we deployed the mobile application on more than 10 users' phones, who navigated in different areas using their cars. We applied the gradient-boosted trees algorithm to predict signal strength degradation in different areas. Our model has shown some interesting results after studying and analyzing the collected data, based on data mining algorithms. We also evaluated our model's ability to predict the zone's structure according to the strength of the degradation signal.