A Comprehensive Validation Methodology for Trajectory Pattern Mining of GPS Data

The increasing pervasiveness of mobile devices along with the use of technologies like GPS, Wifi networks, RFID, etc., allowed for the collections of large amounts of movement data. This amount of information can be analyzed to extract knowledge, i.e. patterns, rules and regularities, from the user trajectories. Currently, most of the approaches for mobility pattern mining are based on a combination of density-based clustering and sequential pattern mining concepts. Specifically, the core strategy consists of (i) discovering regions of interest, and (ii) extracting trajectory patterns from those regions. Nevertheless, the existing literature lacks of a rigorous validation approach aimed at assessing accuracy and quality of the discovered regions of interests and trajectory patterns. Validating the adequacy of the mined trajectory rules to the real mobility patterns represents a crucial aspect. In this paper we present a novel comprehensive validation methodology for assessing the quality of discovered trajectory patterns, i.e., evaluating how the discovered knowledge model fits to the input data it is discovered from. A detailed experimental evaluation proves the effectiveness of the proposed methodology to assess the accuracy and quality of both dense regions and trajectory patterns discovered.