Users’ visiting patterns to POIs (Points-Of-Interest) varied with regard to the users’ familiarity with their visited areas. For instance, users visit tourist sites in unfamiliar cities rather than in their familiar home city. Previous studies have shown that familiarity can improve POI recommendation performance. However, such studies have focused on the differences between home and other cities, and not among small urban neighborhoods in the same city where user activities frequently occur. Applying the studies directly to the areas is difficult because simple distance-based familiarity measures, or visit-pattern differences represented on topics, groups of POIs that share common functions such as Arts, French restaurants, are too coarse for capturing the differences observed among different areas. In the urban neighborhoods in the same city, user visit-pattern differences originate from more precise POI levels. In order to extend the previously proposed familiarity-aware POI recommendation to be adopted in different areas in the same city, we propose a method that employs visit-frequency-based familiarity and precise POI level of visit-pattern differentiation. In experiments on real LBSN data consists of over 800,000 check-ins for three cities: NYC, LA, and Tokyo, our proposed method outperforms state-of-the-art methods by 0.05 to 0.06 in Recall@20 metric.
- Location-based social network
- POI recommendation
- Urban neighborhoods
ASJC Scopus subject areas
- Computer Science(all)