2019 Volume 75 Issue 5 Pages I_507-I_517
The purpose of this study is to identify the visited places combination pattern of foreign travelers visiting Japan. The Hierarchical Pachinko Allocation Model which extracts the combination patterns of visited places is introduced at first to clarify the difference from the Latent Dirichlet Allocation Model. The hPAM, which is regarded as one of the machine learning methods, is able to indicate the probabilistic derivation process of dividing each visited pattern into topics. The hPAM also has a feature that can consider the relevance between topics as a hierarchical structure. In the analysis, using the “Consumption Trend Survey for Foreigners Visiting Japan” conducted by the Japan Tourism Agency, the combination patterns of visited places by the hPAM based on the presence or absence of actual point-by-point visited data are classified into 10 super topics and 35 sub topics under the hierarchy of these topics. Using these outputs, the relations between topics and personal attributes, which are nationality and visit frequency, are finally identified.