부산대학교 도서관

This study proposes a new method for categorizing and visualizing the characteristics of user evaluation tendencies via customer satisfaction data from hotel reservation sites. We focused on the fact that there are individual differences in the way customers evaluate items, and that there are items for which the evaluation is lenient and others for which it is not. We proposed a new method for calculating feature quantities; furthermore, we proposed a user clustering and visualization method using SHAP and UMAP for feature values and analyzed the data. The proposed method enables us to visually understand the features of the data by comparing the relationship between the SHAP values and evaluation trends on a low-dimensional map. The data were based on customer data provided by Rakuten, a major IT company in Japan.