Abstract: A system and method for automatically assigning visits to partially observable location data streams to maintain a geo-spatial model of a real world are provided. The method includes identifying a subset of a plurality of data streams that have a power-law characteristic in a time dimension or spatial dimension associated with the activity of the plurality of entities, modelling an activity of the plurality of entities to determine a region of uncertainty, obtaining a filtered activity of the entities, clustering the time dimension and the spatial dimension of the filtered activity using a stay points clustering method to generate at least one valid data stream, generating a confidence score for the at least one valid data stream, and assigning a visit of the plurality of entities to a point of interest (POI) based on the confidence score of the at least one valid data stream.