Keywords:Bayesian Network, Clustering, User modeling
One of the most effective ways to estimate user types is to administer a questionnaire. In most cases, a questionnaire has multiple questions, and requires considerable time to complete.To make the process time-efficient, we proposed an adaptive questionnaire system based on a Bayesian network in our previous study. Under this system, a question evaluation function is used to ensure that subsequent questions are adapted to the respondents’ answers to preceding questions. Afterward, the respondents are categorized according to user types. In this study, we modify the question evaluation function in the previous system to enable it to estimate the answers to unanswered questions based on the estimated user type. Experimental results show that although the proposed method is inferior to the method that directly estimates answers without using segments, it is superior to the original method that is specialized for user type estimation.