Analyzing regional characteristics of living activities of elderly people from large survey data with probabilistic latent spatial semantic structure modeling

摘要

This paper analyzes a questionnaire survey data on elderly people in order to investigate regional characteristics of their living activities. For the purpose, we use Probabilistic Latent Spatial Semantic (PLSS) Modeling, which is integrated the two methods: probabilistic latent semantic analysis (pLSA) and Bayesian network (BN). First, we aggregate each individual's survey record by postal code; Second, we find characteristics of the region by pLSA; Third, we use BN to clarify factors of this regional disparity. From the study, we are able to identify critical information to support decisions for a manager in a local government: i) The responses of the elderly are correlated with their resident areas; and ii) The regional disparity of social network will improve by neighborhood facilities. Such information will be of use for designing the super-aged society in the near future.