A diversity personalization approach towards recommending POIs for Jeju island

Abstract

In this paper, we propose a diversity personalization approach: we find out how important each user considers diversity of a travel destination when choosing where to visit. Then, we provide recommendations on tourist attractions by combining the score of the personalized diversity, predicted rating score and popularity of POI. We crawled TripAdvisor and Naver data to evaluate the proposed method. Experimental results show that the proposed method shows meaningful improvements in Recall, nDCG, and MRR in terms of top-1, top-2, and top-3 recommendations compared to several baselines.