Segmenting port customer based on convolution kernels and k-aggregate clustering

With the aggravation of port economic competition, port enterprise managers pay special attention on port customer relationship management (CRM). Customer segmentation as a basis of CRM has important research meanings. In this paper, according to the characteristics of port customer data, a tree structure of data organization was obtained, reorganized the port data in 219 customer trees, introduced the concept of convolution kernels, a convolution tree kernel was defined; soon afterwards designed kernel k-aggregate clustering algorithm that suitable for tree structure data, combined the actual situation of port, segmented the customers of port into five groups; at last gained a good segmentation result verified by MATLAB data process tool.