Mixed-initiative synthesized learning approach for web-based CRM

The objective of this approach is to hierarchically segment data sources into clusters, automatically label the features of the clusters, discover the characteristics of normal, defected and possibly defected clusters of customers and provide clues for gaining customer retention.

Abstract

The issue of customer relationship management has emerged rapidly. Customers have become one of the most important considerations to new companies being built. Accordingly, customer retention is a very important topic. In this paper, we present a mixed-initiative synthesized learning approach for better understanding of customers and the provision of clues for improving customer relationships based on different sources of web customer data. The approach is a combination of hierarchical automatic labeling SOM, decision tree, cross-class analysis, and human tacit experience. The objective of this approach is to hierarchically segment data sources into clusters, automatically label the features of the clusters, discover the characteristics of normal, defected and possibly defected clusters of customers, and provide clues for gaining customer retention.