Comparison of K-means, Normal Mixtures and Probabilistic-D Clustering for B2B Segmentation
Satish Garla, Goutam Chakraborty, Oklahoma State University, and Gary Gaeth, University of Iowa
Cluster Analysis is a popular technique used by organizations for market segmentation. Clustering splits customers in a market into groups such that the customers within a group are similar and customers between the groups are dissimilar. Several clustering algorithms were suggested in the literature based on a variety of similarity measures. This poster describes a comparative study of three clustering methods (K-means, Normal Mixtures and Probabilistic-D) for segment profiling of customers in a business-to-business (B2B) market. Data collected from a survey conducted by a supplier of hydraulic and pneumatic products was used in this study. Ten variables that measure customers’ perceptions of important attributes in selecting a supplier were used for clustering. The results from each method are evaluated based on cluster purity and cluster profiles. SAS® Enterprise Miner is used for probabilistic-D clustering and for profiling clusters while JMP® Pro 9 is used for K-Means and Normal Mixtures.