Currently, the increase of location aware services and network management has driven the demand for user location estimation schemes, although it is not usually available to operators. Moreover, commercial networks have limited access to specific user related metrics. In general, solutions with Machine Learning (ML) have reached high precisions, but only in a trained scenario, and with difficulties in predicting unseen areas. The approach proposed here solves the above limitation by a reference coordinate conversion, to obtain relative polar positions which create scenario agnostic models, and whose performance is demonstrated using a dataset recollected from a commercial mobile network.
Buscar en Dimension
