A visualization method of third-order tensor for knowledge extraction from questionnaire data

This paper presents a new visualization method based on Higher Order Singular Value Decomposition(HOSVD). This method enables us to select any pairs of vectors from loading matrices, and visualizes features of loading vectors in the form of hi plots of respondents, questions and objects. An experiment is carried out using artificial data. The results shows that the proposed method can visualize the interrelationships between features of selected loading vectors and we can find respondent groups who marked uniquely on particular objects and questions. Some groups are unable to be found by the conventional PCA.