Using Semantic Similarity to Improve Information Discovery in Spatial Data Infrastructure

Trajectory data are normally generated as sample points, which are very difficult to understand and to analyze because they are often collected with no semantic information. Several studies have been developed for trajectory data analysis. Recently, a new model was designed to reason over trajectories as stops and moves, where stops are the important parts of trajectories. Based on this work, different methods have been developed to instantiate this model, based on different characteristics like speed and direction, aiming to give more semantics to trajectories. In this work we go one step forward to existing works that compute stops of trajectories. We evaluate the behavior of a trajectory considering first its geometric properties like velocity and direction change, and then, based on this analysis we propose to use domain knowledge that describes some characteristics of the application domain to infer the goal of the stops. To validate the proposed method we present some experiments over real trajectory data.