Clustering Algorithm on Spatiotemporal Trajectories

Now days, there exist huge data resources such as transaction database, relational database, data warehouse and other data systems all contain large amount of data. However, in these data it is very hard to find useful information like looking for something in an ocean. More difficult problem is to find inner rules and relationship and how to predict the trend of the future based on existing data. Therefore, it has become a problem in computer science and technology to capture useful information out of resources; data mining gives us solution of this problem. Clustering which is function of data mining can be used as single step in data mining algorithm. Since moving object generates trajectory that has two properties, spatial is a land (space) property and temporal is a time property. Perhaps, there are three problems how to combine both properties of moving object in order to extract informationƒ How fast an algorithm finds similar patternsƒ And how many more clusters are generated by an algorithmƒ This paper carries deep research on clustering and addresses above problems by proposing clustering algorithm on spatiotemporal trajectories. The results show that, our proposed clustering algorithm has high efficiency than prior most frequently used algorithm DBSCAN for trajectories.