The HPCLUS Procedure
Overview: HPCLUS Procedure
The HPCLUS procedure is a high-performance procedure that performs clustering, which is a common step in data exploration. You can use the HPCLUS procedure to perform clustering and scoring in parallel by making full use of multicore computing environments.
The HPCLUS procedure performs a cluster analysis on the basis of distances that are computed from one or more quantitative variables. The observations are divided into clusters such that every observation belongs to one and only one cluster.
The HPCLUS procedure uses the k-means algorithm for clustering numeric interval input variables and uses the k-modes algorithm for clustering nominal input variables. The HPCLUS procedure uses only numeric interval variables or only nominal variables to perform clustering. It does not perform clustering for mixed levels of input variables.
The HPCLUS procedure uses the least squares (
) estimation in the k-means clustering method to compute the cluster centroids. In this method, each iteration reduces the criterion (for example, the least squared criterion for Euclidean distance) until convergence is achieved or the maximum iteration number is reached.
PROC HPCLUS produces brief summaries of cluster analysis in two output data sets:
The OUT= data set is produced by the SCORE statement. This data set contains the cluster membership and distance variables for each observation in the input data set. It can be used for more detailed examination of the clusters.
The OUTSTAT= data set is produced by the PROC HPCLUS statement. This data set can be used for more detailed examination of between-cluster statistics.