Machine Learning and Radio Planning in the Design and Optimization of Wireless Networks

This paper describes the incorporation of Machine Learning in the process of designing and optimization of a Long Term Evolution network in an urban parish of Quito city performed with the help of a radio planning tool. Potential subscribers behavior within this area was extracted from an online survey. Two Machine Learning methods were applied to this data: K-Nearest Neighbors and K-Means Clustering. The former was used to identified the dissatisfied users whilst the latter to classify network users into adequate categories that were included in the design. Details of the environment were also taken account, by adding several clutter maps that include geographical information, roads, buildings and traffic patterns with the aid of a Radio Planning software that provides the propagation models and the means required for the accurate modelling, and eventually the optimization of the deployed network.