Multi-Relational Approach for Spatial Clustering

A growing attention has been paid to spatial data mining and knowledge discovery. Spatial clustering is the process of grouping a set of objects into classes or cluster that objects within a cluster have high similarity in comparison to one another, but are dissimilar to objects in other clusters. As to the data, several clustering algorithms assume that observations are represented as tuples of a single database relation. This “single table assumption� prevents the consideration of relationships between observations as well as the analysis of data which are logically modeled through several database tables. To overcome these limitations, many researchers have recently started investigating the Multi-Relational Data Mining (MRDM) approach to data analysis. In some sense, graph-based partitioning methods also represent a particular class of multi-relational clustering algorithms. Moving towards a structured concept of spatial object forces to arrange data in different tables, related each others, where each table describes either features corresponding to entire areas or primary units or relations between them.