Tampakis, P.P.TampakisPelekis, N.N.PelekisTheodoridis, Y.Y.TheodoridisAndrienko, NataliaNataliaAndrienkoAndrienko, GennadyGennadyAndrienkoFuchs, GeorgGeorgFuchs2022-03-142022-03-142018https://publica.fraunhofer.de/handle/publica/40737410.1109/ICDE.2018.00181In this paper, we present an efficient in-DBMS framework for progressive time-aware sub-trajectory cluster analysis. In particular, we address two variants of the problem: (a) spatiotemporal sub-trajectory clustering and (b) index-based time-aware clustering at querying environment. Our approach for (a) relies on a two-phase process: a voting-and-segmentation phase followed by a sampling-and-clustering phase. Regarding (b), we organize data into partitions that correspond to groups of sub-trajectories, which are incrementally maintained in a hierarchical structure. Both approaches have been implemented in Hermes@PostgreSQL, a real Moving Object Database engine built on top of PostgreSQL, enabling users to perform progressive cluster analysis via simple SQL. The framework is also extended with a Visual Analytics (VA) tool to facilitate real world analysis.en005006629conference paper