CIFEF: Combining Implicit and Explicit Features for Friendship Inference in Location-Based Social Networks.

Cheng He,Chao Peng,Na Li,Xiang Chen,Zhengfeng Yang,Zhenhao Hu

DOI: https://doi.org/10.1007/978-3-030-55393-7_16

2020-01-01

Abstract:With the increasing popularity of location-based social networks (LBSNs), users can share their check-in location information more easily. One of the most active problems in LBSNs is friendship inference based on their rich check-in data. Previous studies are mainly based on co-occurrences of two users, however, a large number of user pairs have no co-occurrence, which weakens the performance of previous proposed methods. In this paper, we propose a method CIFEF that C ombines the Implicit Features and a Explicit Feature for friendship inference. Specifically, based on whether a user has different trajectory patterns on weekdays and weekends, we take the embedding technique to learn implicit weekdays’ trajectory features and weekends’ trajectory features from their check-in trajectory sequences, respectively, which can work effectively even for user pairs with no co-occurrence. Moreover, we propose a new explicit feature to capture the explicit information of user pairs who have common locations. Extensive experiments on two real-world LBSNs datasets show that our proposed method CIFEF can outperform six state-of-the-art methods.