Spatio-temporal data streaming with affinity propagation

Spatio-temporal data stream clustering is a growing research field due to the vast amount of continuous georeferenced data streams being generated by IoT devices. Carnein and Trautmann (2019) provide an extensive review on stream clustering algorithms, outlining an overall strategy that is based on a two-phase clustering approach; having an online phase which uses a time window model to capture the data streams and then computing micro-clusters (i.e. preliminary clusters within each time window). The second phase is carried out offline as the micro-clusters are re-clustered to generate the macro-clusters after the entire stream data is processed. Distance-based algorithms such as CluStream using the pyramidal time window model (Aggarwal et al. 2003) and DenStream using the damped time window model (Cao et al. 2006), are widespread approaches used in stream clustering.