Abstract:

This paper describes an application of automated website synthesis and uncertainty reasoning for customer preference management (CPM), particularly in the context of web-based customer relationship management (CRM). Stated preference elicited by conjoint analysis and revealed preference acquired by pattern analysis of the interaction history with customers are useful for CRM agents (software agents) to make decisions about further actions to meet probable needs of the same or similar customers. As available preference information is never sufficient, CRM agents must reason under uncertainty. Use of Bayesian networks and algorithms of partially observable Markov decision processes (POMDP) is proposed in reasoning about customer preference under uncertainty. We are also developing a technique of automated website synthesis by using computation logics and Semantic Web languages to facilitate the interactions with different customers.