A Novel Moving Object Trajectories Clustering Approach for Very Large Datasets

DOI

10.2991/iccsee.2013.217How to use a DOI?

Keywords

Moving Object Trajectories, clustering, algorithm

Abstract

Witnessing a rapid and continuous diffusion of mobile devices such as on-board GPS navigators, smartphones and tablet computers, we also facing the challenge that how to effectively and efficiently cluster the trajectories derived from these devices. Such an algorithm can benefit a range of applications and services, including intelligent transportation systems, personalized route planner, smart sports and predictionbased social network, etc. Coping with moving object trajectory clustering has long been an important research direction on moving object pattern mining, but still remains two difficulties to deal with. First, how to define the similarity between two trajectories. Second, how to cluster a large number of trajectories efficiently. To define the similarity, a key point is to choose appropriate granularity. If fine-grained policy is adopted, we can expect a more precise result. However, it requires a lot of computations. On the other hand, if coarse-grained strategy is used, we can quickly cluster very large number of trajectories. However, the results are not as good as fine-grained algorithm. Hence, it’s a dilemma. To cluster a large number of trajectories, a lot of algorithms are proposed. Some of them are focusing on finding moving objects which move close to each other for a time duration, such as: swarm. Some of them employ the notion of density connection in order to enable the formulation of arbitrary shapes of groups, such as: convoy. In this paper, we aim at find the similarity between the trajectory sets, where each set is generated by a moving object. In this way, we can explore the personalized patterns which are more crucial to a lot of real-world applications as mentioned above. First, we define the similarity between two moving object trajectories. Then, we give an algorithm to cluster trajectory sets. Finally, intensive experiments are conducted and the results prove both the effectiveness and efficiency of our algorithm.

Copyright

© 2013, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Jian Dai
PY  - 2013/03
DA  - 2013/03
TI  - A Novel Moving Object Trajectories Clustering Approach for Very Large Datasets
BT  - Proceedings of the 2nd International Conference on Computer Science and Electronics Engineering (ICCSEE 2013)
PB  - Atlantis Press
SP  - 863
EP  - 866
SN  - 1951-6851
UR  - https://doi.org/10.2991/iccsee.2013.217
DO  - 10.2991/iccsee.2013.217
ID  - Dai2013/03
ER  -