Clustering is a technique for grouping a large number of objects and generates clusters, which are used to summarize the original dataset. There have been a lot of clustering algorithms proposed in data mining [25, 56]. In the following, we introduce some clustering techniques for moving object databases.

11.5.1 Continual Maintenance of Moving Clusters

Li et al. [36] proposed a real-time and adaptive cluster maintenance method for moving points. The approach is based on the notion of micro-clusters. A micro-cluster is a small-sized cluster consisting of nearby objects. After the generation of micro-clusters, some different clustering algorithms can be applied to the micro-clusters by treating each micro-cluster as if it were an individual entity. The idea of micro-clusters was initially proposed in BIRCH [69]. The method in Ref. [36] generates moving micro-clusters from the target moving objects, and then global clusters are generated using the micro-clusters. Since moving objects change positions and directions, the method maintains clusters adaptively.

The merging and partitioning processes for micro-clusters are performed using clustering features, which summarize the clusters. A clustering feature for a (micro-) cluster Ci is defined as


where ti is the cluster generation time, ni the number of elements (ni = |Ci|), and sxi (syi) the sum of the ...

Get Mobile Intelligence now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.