An approach for improving K-means algorithm on market segmentation
Haibo Wang, Da Huo, Jun Huang, Yaquan Xu, Lixia Yan, Wei Sun, Xianglu Li · 2010
The K-means algorithm is among the most popular clustering methods that group observations with similar characteristics or features together. It is widely used in many marketing applications, especially in cluster-based market segmentation. The K-means algorithm is implemented by different commercial software, such as SAS, SPSS and MATLAB, as a standard clustering function/tool. This note compares the performances of K-means algorithm implemented by three software. This note describes the potential shortcomings of the K-means algorithm implementation within the software, and proposes improvement approaches for the K-means algorithms by using silhouette coefficient.