The functions Q.PredictClusterID, Q.PredictClusConf and Q.PredictClusScore are based on a mining technique called clustering. This data mining technique discovers sets of rows with common characteristics, or clusters, so that the rows are possibly homogeneous inside a cluster and possibly heterogeneous between two clusters.
This technique is typically used for customer segmentation and profiling. It divides customers into groups, based on parameters such as usage patterns, cross-selling opportunities, and demographic data.