Treffer: Big mobility data analytics: algorithms and techniques for efficient trajectory clustering ; Αναλυτικές μέθοδοι μεγάλων όγκων δεδομένων κίνησης: αλγόριθμοι και τεχνικές για αποτελεσματική ομαδοποίηση τροχιών

The unprecedented rate of trajectory data generation that has been observed during the recent years, caused by the proliferation of GPS-enabled devices, poses new challenges in terms of storage, querying, analytics and knowledge extraction from mobility data. One of these challenges is cluster analysis, which aims at identifying clusters of moving objects according to the similarity degree of their movement. Discovering clusters of moving objects is an important operation when trying to extract knowledge out of mobility data, since by doing so, the underlying hidden patterns of collective behavior can be unveiled. What is even more challenging is treating knowledge discovery techniques, such as cluster analysis, as an integral part of a real DMBS, which can turn out to be practical and useful in real-world application scenarios, where issues like the ease of use (e.g., via a simple SQL interface) are taken into consideration. Furthermore, the support of incremental and progressive cluster analysis in the context of dynamic applications is of great interest, where (i) new trajectories arrive at frequent rates, and (ii) the analysis is performed over different portions of the dataset, and this might be repeated several times per analysis task. However, performing such “expensive” operations over immense volumes of data in a centralized way is far from straightforward, which in turn calls for parallel and distributed algorithms to address the scalability requirements posed by the Big Data era. The bottleneck of performing “expensive” operations, such as cluster analysis, is the underlying join query. Joining trajectory datasets is not only the cornerstone of various trajectory cluster analysis methods, but it is also a significant operation in mobility data analytics with a wide range of applications, such as carpooling, suspicious movement discovery, etc. In this thesis, we aim to address the above challenges. Towards this direction, we propose a novel in-DBMS Sampling-based Sub Trajectory Clustering algorithm, ...