Research Area:  Machine Learning
Group recommendation has attracted researchers attention in various domains, specifically such approaches utilizing location-based social networks (LBSNs). However, point of interest (POI) group recommendation faces the challenge of aggregating diverse user preferences, while group members have different influences on the final decision of the group. Besides, the recommendation of spatial items is different from non-spatial items and the unique features of the spatial items such as distance must be considered in the recommendation. In this paper, a POI group recommendation method is proposed to tackle this problem. User influence is modeled fuzzy and taken into account the difference of users personality and their preferences when are alone or in a group, by using historical check-in data in LBSNs and in terms of category, distance and time. The proposed method is integrated with the weighted average aggregation to improve the efficiency of the POI group recommendation. Experimental results in a real dataset show improvement in the accuracy of POI group recommendations in varying sizes of groups. The results also get better when the user influence is calculated using the fuzzy approach. Besides, studying user behavior differences to choose the place to visit when alone or in a group shows that i) the flexibility of users in distance is less than time and category. It is also in the category less than time. ii) Time has a greater range of behavioral change than distance and category. iii) Users who actively participate in group decision making have a more significant number of visits in groups than when they are alone.
Author(s) Name:  Zahra Bahari Sojahrood. Mohammad Taleai
Journal name:  Expert Systems with Applications
Publisher name:  ELSEVIER
Volume Information:  Volume 171, 1 June 2021, 114593
Paper Link:   https://www.sciencedirect.com/science/article/abs/pii/S0957417421000348