TMC-pattern: holistic trajectory extraction, modeling and mining

Abstract

Mobility data is Big Data. Modeling such raw big location data is quite challenging in terms of quality and runtime efficiency. Mobility data emanating from smart phones and other pervasive devices consists of a combination of spatio-temporal dimensions, as well as some additional contextual dimensions that may range from social network activities, diseases to telephone calls. However, most existing trajectory models focus only on the spatio-temporal dimensions of mobility data and their regions of interest depict only the popularity of a place. In this paper, we propose a novel trajectory model called Time Mobility Context Correlation Pattern (TMC-Pattern), which considers a wide variety of dimensions and utilizes subspace clustering to find contextual regions of interest. In addition, our proposed TMC-Pattern rigorously captures and embeds infrastructural, human, social and behavioral patterns into the trajectory model. We show theoretically and experimentally, how TMC-Pattern can be used for Frequent Location Sequence Mining and Location Prediction with real datasets.