| dc.description | The wide use of mobile devices such as GPS, as well as the popularization of social media, has led to the generation of large amounts of movement data, called trajectories of moving objects. Trajectory data analysis and mining has become very important because of the variety of information that may be extracted/inferred from these data, such as the daily habits or the profile of individuals. Because of the complexity of the data, they must be analyzed not only from the spatial and temporal characteristics, but any other semantics that may be related to the data. Behind the large amount of information available about movement, trajectories may be analyzed from multiple points of view, that we call multiple aspect trajectories. Similarity measures are widely employed for trajectory data analysis and have a large impact on the analysis outcomes. Most existing works for trajectory similarity are limited to the space and time dimensions of trajectories, and only a few analyze some semantic characteristics of trajectories. Works such as LCSS, EDR and MD-DTW are very rigid and limited to the order of the trajectory points, and two trajectories are considered similar if they match on all dimensions. On the other hand, works such as MSM are too flexible, considering two trajectories as similar if they match in any dimension. In this work, we define the concept of multiple-aspect trajectory, proposing the use of several attributes regarding different aspects related to movement. We propose MUITAS, a novel similarity measure for multiple-aspect trajectory similarity analysis, which overcomes the described limitations of previous works. MUITAS is evaluated on a toy example and over a real dataset of user check-ins on a social network containing different aspects related to movement. The results show that MUITAS is more accurate than existing similarity measures for analyzing multiple-aspect trajectories, in addition to allowing the analysis of trajectories in ways not explored before. | |
