Effective partial-answered data acquisition method based on Probabilistic Latent Semantic Analysis for consumer behavior modeling according to occasion

Abstract

Taking a questionnaire is important for understanding the user, but it is very costly to the user. In the previous study, we proposed a method that creates full-answered data by merging multiple partial-answered data. However, there is a decreased accuracy problem of created survey data according to increase in the number of questionnaire questions. In this study, we propose a new partial-answered collecting method according to user’s characteristics and a modified merging method. The experimental result showed that the proposed methods can create full-answered data similar to the original questionnaire from the point of view of user segment.