Thesis (Ph.D.) - Cairo University - Faculty of Computers and Artificial Intelligence - Department of Information System

The fast advancement that we are witnessing today in mobile computing techniques has generated massive spatiotemporal data. Mining spatiotemporal data and especially outlier detection in trajectory data is a crucial and challenging topic as it can be used in a wide range of applications, including transportation management, public safety, urban planning, and environment monitoring. An outlier (anomaly) trajectory is a trajectory that has different characteristics than normal trajectories. In our research, we present the CB-TOD algorithm to detect outlier sub-trajectories and outlier trajectories by utilizing a clustering-based methodology. In the CB-TOD algorithm, the computational time is reduced decreasing the size of the trajectories dataset and representing each trajectory with the summary set of line segments that are sufficient to define the trajectory behavior without missing the basic motion information. After that, similar line segments based on the distance are grouped into a cluster. After clustering, for each trajectory, we distinguish the cluster that has the smallest number of segments and neighbors.This cluster is marked as an outlier cluster for this trajectory and accordingly, the line segments included in this detected cluster are classified as outlier segments. Moreover, a trajectory that contains a considerable number of outlying partitions is identified as an outlier trajectory

Issued also as CD