Abstract
Recently, diversification of eating habits has been advancing in Japan. In such a situation, issues have been raised on the eating habits of younger generation, especially university students. The purpose of our study is to provide dietary life assistance to the university students by developing a system to locate students with unbalanced diet. Meal purchase patterns of university students are analyzed by utilizing purchasing history of the meal prepaid card users. The data are provided by a university co-operative association who runs university cafeterias. The number of the users analyzed are 2,458 (male 1,591 and female 867). The procedure of the analysis is as follows: k-means clustering is applied to the combination of food to generate 22 food clusters. Next PLSA is applied for clustering the customers to generate 8 customer clusters. For these two clusters, Bayesian network is applied to visualize correlation among factors. Based on this visualization and extraction of characteristics through clustering, customer clusters with eating habit problems are detected. Specific proposals for dietary life improvement are proposed to these clusters.
