A Kernel Aggregate Clustering Approach for Mixed Data Set and Its Application in Customer Segmentation

There are lots of categorical valued and mixed numeric and categorical valued data in the practical application, and the traditional clustering methods can't analyses this kind of data very well. Aiming at the clustering of the mixed valued data, and a advanced kernel k-aggregate clustering algorithm is presented by combining the clustering analysis with kernel-based method. In this algorithm, the aggregate function (i.e. maximum entropy function) to approximate the maximum function and the categorical valued attributes decompose (CVAD) are applied in order to effectively define the computing scheme and the distance for categorical valued data. Like the fuzzy k-prototypes algorithm, the algorithm is also soft clustering, but it is more easier and insensitive to the selection of the aggregate parameter than the fuzzy one. So the algorithm is applied to the customer segmentation and gets a good clustering result which provides the managers guidance and evidence of different marketing strategies for corresponding subdivided markets. And in the clustering process the best clustering number is chosen by the significance test on five selected attributes.