Effective fine-grained location prediction based on user check-in pattern in LBSNs. (15th April 2018)

• Record Type: Journal Article
• Authors: Cao, Jiuxin; Xu, Shuai; Zhu, Xuelin; Lv, Renjun; Liu, Bo
• Is Part Of: Journal of network and computer applications. Volume 108(2018)
• Page Start: 64
• Page End: 75
• Publication Date: 2018-04-15
• Subjects: Location-Based Social Networks -- Location prediction -- Check-in pattern analysis -- Scoring model -- Classification model; Microcomputers -- Periodicals; Computer networks -- Periodicals; Application software -- Periodicals; Micro-ordinateurs -- Périodiques; Réseaux d'ordinateurs -- Périodiques; Logiciels d'application -- Périodiques; Application software; Computer networks; Microcomputers; Periodicals; 004.05; 004
• DOI: 10.1016/j.jnca.2018.02.007
• Languages: English
• ISSNs: 1084-8045

Abstract

Abstract: Location-Based Social Networks (LBSNs) have built bridges between virtual space and real-world mobility in recent years. The massive check-in data generated in LBSNs makes it possible to predict users' future check-in location, which has proved meaningful for e-commerce developments. Existing studies mainly focus on predicting the next check-in location with a coarse granularity, which only shows limited performance in practical scenarios. In this paper, we propose a comprehensive approach based on user check-in pattern to predict users' future check-in location at any fine-grained time in LBSNs. Firstly, users' check-in pattern involving time periodicity, global popularity and personal preference are analyzed. Secondly, we extract multiple features related to user check-in pattern and explore the predictive power of each individual feature. Thirdly, a set of features are combined into a supervised scoring model and a classification model respectively for predicting user's check-in location at a fine-grained time in the future. Finally, extensive experiments on three real-world Foursquare datasets are carefully designed to verify the effectiveness of the proposed approach. Experimental results show that our approach outperforms both baseline methods and state-of-the-art methods on various evaluation metrics. Highlights: Multiple features related to user check-in pattern are extracted. The predictive power of each individual feature is widely explored. A set of features are combined into a supervised scoring model to improve prediction performance. A supervised classification model is proposed to simplify the location prediction problem. The proposed scoring model and classification model are combined to strengthen the predictive performance.