next up previous contents
Next: Balancing Up: Relationship-based Clustering and Visualization Previous: Domain Specific Features and   Contents


In this section, we propose OPOSSUM (Optimal Partitioning of Sparse Similarities Using Metis), a similarity-based clustering technique particularly tailored to market-basket data. OPOSSUM differs from other graph-based clustering techniques by application-driven balancing of clusters, non-metric similarity measures, and visualization driven heuristics for finding an appropriate $ k$.


Alexander Strehl 2002-05-03