Improving Telco Data Positioning Using Infrastructure-Aware Mapping

Today, in the big data era, geolocation data is one of the most important kinds of data and can be extremely useful for both academic and business purposes. Such data is usually collected in the form of GPS points. Telecommunication-based geolocation data can be a promising alternative to GPS, since such data provides broader coverage. However, telco data has significantly lower accuracy. This work aims to enhance spatial precision of telco data received from cell tower connections. To reach this goal, we propose several approaches, encompassing both supervised and unsupervised methods, along with using open-source map data (Overture Maps). Our main result is the Q2B (Quadkey to Buildings) method, which operates with quadkey spatial indexing and building-level metadata to associate telco-geolocation points with buildings in a probabilistic manner. In this work we also explore other approaches, such as uniform data redistribution, spatial clusters deconstruction and building an uncertainty distribution. Experimental evaluations of our results confirm that our methods are capable of improving geolocation accuracy of telco data as well as providing meaningful insights into events and user behavior.