CluPlaR: A Framework for Clustering Spatio-Temporal Geographic Positions

Fitness trackers, smart watches and also smartphones utilize different types of sensors to track physical activities of an active person. These measurements are often combined with geographical locations at which the sensors capture data which are then transformed to qualitative or quantitative measurements. However, for the later analysis the individual position data are often too fine-grained and, hence, need to be aggregated to locations ranging from places similar to the real world to larger regions, such as districts and whole cities. The CluPlaR (Clustering based Place Recognition) framework enables the aggregation of position data to places and their chronological visits. In this paper, we provide its conceptual foundation and corresponding algorithm allowing to aggregate chronologically collected geographic position data as they have often produced by today’s activity sensors. In detail, we describe the algorithm and corresponding configuration parameters and examine their influence on the aggregation result. The evaluation results show that the algorithm can be used to efficiently aggregate position data, but also that the result depends on the chosen aggregation parameters. In addition to the application of the CluPlaR algorithm in real application scenarios, future work should focus on the automated determination of the aggregation parameters.