Abstract
We extracted latent topics of each participant for health checks data based on the medical interview items to illustrate validity of medical interview data for evaluation of the health conditions, which the stochastic topic model is defined by latent Dirichlet allocation. Clusters of the participants with the same topics were compared with the clusters based on modularity of the network graph. We extracted 30 latent topics from 4,384 participant's data with 270 items. Comparisons between participants with lifestyle-related disease items and participants with normal lifestyle items showed significant differences for GLU (p<0.05). From these results, the medical interview data of each participant proven to be useful data for comprehend the condition as the same as chief complaint and taking history for diagnosis.
