A Method of Expanding Photo-Based User Preference Profile for the Recommendation of Sightseeing Places.

This study proposes a method of expanding photo-based users’ preference profiles for the recommendation of sightseeing places. In order to recommend the desired places for the new users, to find the places with unexperienced sceneries and interesting activities. We found some difficulties in conventional studies, involving that the domain of preference analysis is too closed to the personal photo archive. Hence, in this study, we improve to expand the photo-based profiles by predicting users’ unexperienced categories of sightseeing places from the photo sets in the photo-based profiles, using a deep neural network, which learning with a sequence pattern of photos taken by different users. Because of the expanded photo-based users’ profiles including visual information of unexperienced categories, it allows the recommendation system to suggest the impressive and beautiful spots where the new users who never has the experiences in that places. In the experiment, we evaluate the feasibility of our proposed method that it can expand the photo-based user profiles for guiding the sightseeing places that is expected to influence the user before the user has interests.