Behavior‐based POI recommendation for small groups in location‐based social networks.

In point‐of‐interest (POI) group recommender systems, suggesting the most suitable visit places for a heterogeneous group in which members are far from each other would be difficult in terms of finding a consensus among group members. In this article, we argue that the geographical proximity of POIs to users' locations has a notable influence on group decisions to visit the POI and their check‐in behavior. We develop a new geographical model based on the check‐in behavior of the group in location‐based social networks to improve group recommendations by introducing a new concept called users' spatial ratio in a group vectors and using a 2D kernel density estimation method. To evaluate the performance of the approach, experimental studies were conducted on a real dataset of users and group check‐in activities collected from Foursquare Swarm. In our experiments, the proposed method makes better recommendations compared with other existing approaches. Besides, analyzing group check‐ins shows that in a group visit, the distance between the most visited place by a group member alone and the places visited by the group follows a generalized gamma distribution. Also, the ratio of this distance for the group member to the distance of another member depends on the group size. [ABSTRACT FROM AUTHOR]