Frequent Spatio-Temporal Patterns in Trajectory Data Warehouses

In this paper we present an approach for storing and aggregating spatio-temporal patterns by using a Trajectory Data Warehouse (TDW). In particular, our aim is to allow the\ud analysts to quickly evaluate frequent patterns mined from\ud trajectories of moving objects occurring in a specific spatial\ud zone and during a given temporal interval.\ud We resort to a TDW, based on a data cube model, having\ud spatial and temporal dimensions, discretized according to a\ud hierarchy of regular grids, and whose facts are sets of trajectories which intersect the spatio-temporal cells of the cube.\ud The idea is to enrich such a TDW with a new measure:\ud frequent patterns obtained from a data-mining process on\ud trajectories. As a consequence these patterns can be analysed by the user at various levels of granularity by means of OLAP queries.\ud The research issues discussed in this paper are (1) the extraction/mining of the patterns to be stored in each cell,\ud which requires an adequate projection phase of trajectories\ud before mining; (2) the spatio-temporal aggregation of patterns to answer roll-up queries, which poses many problems due to the holistic nature of the aggregation function