USING PROJECTIONS TO VISUALLY

The difficult task of sifting through thedata generated by the extensive useof computers today—and locatingconcise and interpretable informa-tion within that data—is called knowledge dis-covery in databases (KDD). Data mining refers to one specific step in the KDD process— namely, to the application of algorithms that can extract hidden patterns from data (see the “KDD Process ” sidebar for more information). The data mining technique we focus on in this article is called clustering—partitioning a set of data vectors into clusters and noise such that data vectors within the clusters are similar to each other and that the data items in different clus-ters or noise partitions are not. Recent research has proposed many algorithms for clustering.1 In all these methods, the user must choose th