Opening the black box: interactive hierarchical clustering for multivariate spatial patterns Diansheng Guo, Donna Peuquet, and Mark Gahegan   View Paper (PDF)   Return to Data clustering Abstract Clustering is one of the most important tasks for geographic knowledge discovery. However, existing clustering methods have two severe drawbacks for this purpose. First, spatial clustering methods have so far been mainly focused on searching for patterns within the spatial dimensions (usually 2D or 3D space), while more general-purpose high-dimensional (multivariate) clustering methods have very limited power in recognizing spatial patterns that involve neighbors. Secondly, existing clustering methods tend to be 'closed' and are not geared toward allowing the interaction needed to effectively support a human-led exploratory analysis. The contribution of the research includes three parts. (1) Develop an effective and efficient hierarchical spatial clustering method, which can generate a 1-D spatial cluster ordering that preserves all the hierarchical clusters. (2) Develop a density- and grid-based hierarchical subspace clustering method to effectively identify high-dimensional clusters. The spatial cluster ordering is then integrated with this subspace clustering method to effectively search multivariate spatial patterns. (3) The above two methods are implemented in a fully open and interactive manner and supported by various visualization techniques. This opens up the "black box" of the clustering process for easy understanding, steering, focusing and interpretation. At the end a working demo with US census data is presented. DiSC'03 © 2003 Association for Computing Machinery