Varification of a method that creates virtual full-answered data from partial-answered data based on Probabilistic Latent Model

Abstract

Taking a questionnaire is effective way to model consumer awareness. However hearing a lot of questions is very hard from the point of view of answer accuracy and response rate. One way to solve this problem, we take partial questionnaires that differ among answers and merge some answers so as to contain all answers of questions in the original questionnaire. In this study, we propose a method that is merging the partial answers based on Probabilistic Latent Model. The experimental result shows that the proposed method can create virtual answer data which is similar to original answer data from the viewpoint of latent segments in answer data.