Elsevier

Future Generation Computer Systems

Volume 100, November 2019, Pages 982-993
Future Generation Computer Systems

A novel next new point-of-interest recommendation system based on simulated user travel decision-making process

https://doi.org/10.1016/j.future.2019.05.065Get rights and content

Highlights

  • Our system simulates a user’s travel decision-making process and considers two important factors that influence people’s travel destination choices: preference factors and geographic factors, and integrates them into a unified recommendation process.

  • We leverage tensor to model users’ check-in history based on POI’s classification properties and check-in data’s time information, which overcomes the sparseness of check-in data. Further, we realize the dynamic prediction of users’ preferences.

  • We propose a novel approach to modeling the personalized impact of geographic factors on individual user travel. Specifically, we fit a curve to personally reflect the relationship between the user’s travel distance and travel probability.

  • We conduct comprehensive experiments by comparing our approach with the state-of-the-art techniques over two real-world datasets.

Abstract

POI(point of interest) recommendation systems have been widely investigated in recent years. Currently, most POI recommendation systems only recommend POIs that may be visited by users in the future, and rarely consider next new POI recommendation based on the current time and the current location of a particular user. In fact, next new POI recommendation problem is more challenging for the reason that multiple factors associated with both POIs and users need to be comprehensively incorporated in a unified recommendation system. In this paper, we design a novel and effective next new POI recommendation system. Our system simulates a user’s travel decision-making process by weighing two important factors that affect a user’s travel decision: preference factors and geographic factors. First, we use tensor to model user’s check-in history and dynamically predict user preferences. Then, in order to characterize the influence of geographic factor on individual users, we designed a personalized user similarity calculation method and fitted curves for the target user to reflect the relationship between travel distance and travel probability. Finally, a recommendation list is generated by combining the effects of these two factors on a particular user. Compared with the state-of-the-art POI recommendation approach, the experimental results demonstrate that our system achieves much better performance.

Introduction

In recent years, with the rapid development of mobile Internet technology, positioning technology, wireless sensor technology and the popularity of smart phones, location-based social networks (LBSNs) as shown in Fig. 1 and its application services have developed rapidly. The currently popular LBSNs are Foursquare, Gowalla, Facebook Place, Microsoft GeoLife, Bikely, Flickr, Panomamio, etc. The LBSNs represented by Foursquare, Gowalla, and Facebook Place mainly provide check-in services for POIs. Encourage users to share their favorite POIs with friends in the form of check-in and share their experiences and tips for POIs. The main difference between LBSNs and Online Social Networks (OSN) is that LBSNs add geographical location information. Location data bridges the gap between the physical and digital worlds and enables a deeper understanding of users’ preferences and behaviors. Since users generate a large amount of check-in data in LBSNs, it is possible to recommend the unvisited POIs to users. POI recommendation can help users better understand their city and explore the surrounding environment. Therefore, POI recommendation is of high value to both users and the business owners of POIs.

POI recommendation is one of the most important tasks in LBSNs, which is to provide recommendations of places to users, and has attracted much attention in both research and industry. However, the general POI recommendation system can only recommend POIs globally and ignores the time context that the recommendation result itself should have. Specifically, the general POI recommendation system only considers which POIs users may access in the future, and they cannot predict where users want to go in the next time interval. Furthermore, the general POI recommendation system does not consider whether the POIs recommended to a user have been previously accessed by this user. If the recommendation results are not novel enough, it will seriously affect the user’s experience. In order to provide users with a POI recommendation system more practical and good experience, we have taken next new POI recommendation system as the research object. The next new POI recommendation is an extension of the general POI recommendation. It does not merely recommend the POIs that the user may visit in the future. Next new POI recommendation system takes into account the sequential influence and recommends the POIs that the user may access at the succeeding moment according to various contextual information such as current time and current location. And these POIs were never visited by this user before.

Compared to general POI recommendation, next new POI recommendation is a more difficult task. In order to achieve more accurate and personalized recommendations, next new POI recommendation system will face more difficult challenges as follows:

  • Next new POI recommendation system needs to consider a user’s current location, because the current location is the starting point for the user to travel at the succeeding moment. The POIs that the user may access during the next time interval are closely related to the current location.

  • Next new POI recommendation emphasizes the real time nature of the recommendation system. This requires the recommendation system to be able to focus on the dynamic changes of the user’s preferences in real-time, and give the user satisfactory recommendation results based on the user’s preferences at the current moment and the most recent check-in.

  • Next new POI recommendation system needs to take into account the geographic factors of the recommendation results. Specifically, the recommendation results are POIs that the user may access from the current location in the next time interval. If these POIs are too far away from the user’s current location, there is no feasibility of travel.

Our system is focused on next new POI recommendation. The main contributions of this paper can be summarized as follows:

  • Our system simulates a user’s travel decision-making process and considers two important factors that influence people’s travel destination choices: preference factors and geographic factors, and integrates them into a unified recommendation process.

  • We leverage tensor to model users’ check-in history based on POI’s classification properties and check-in data’s time information, which overcomes the sparseness of check-in data. Further, we realize the dynamic prediction of users’ preferences.

  • We propose a novel approach to modeling the personalized impact of geographic factors on individual user travel. Specifically, we fit a curve to personally reflect the relationship between the user’s travel distance and travel probability. Due to the sparseness of the check-in data of individual users, we established a virtual common access sequence for two users and designed a novel user similarity algorithm to find users with similar historical behavior to the target user, using similar users’ check-in data as supplement of the target user’s data.

  • We conduct comprehensive experiments by comparing our approach with the state-of-the-art techniques over two real-world datasets.

Section snippets

Related work

POI recommendations using user’s check-in history are often influenced by multiple factors such as geography, time, sequence and society. Geographical influences are the most essential feature that distinguishes POI recommendations from traditional recommendations. Since the user’s check-in behavior presents a spatial clustering phenomenon, geographical influence can be modeled by power law distribution [1], Gauss distribution [2], Poisson distribution [3], and kernel density estimation [4], [5]

Problem definition

In order to facilitate our system, we have the following notations:

  • 1.

    U: the set of the entire users.

  • 2.

    F: the set of all the preferences. Each POI has its own category, such as western restaurant, shopping center, park, and so on. Bao et al.’s research [36] suggests that the category of a POI that a user has visited usually implies the user’s travel preferences. In our system, we refer to the categories of POIs as user preferences.

  • 3.

    L: the set of the entire POIs. Each POI lL is represented as l=x,y,f

Our system model

Our system focuses on the next new POI recommendation, which can help users make travel decision. In this paper, the travel decision refers to how a person chooses a travel destination at the succeeding moment. In real life, people’s travel decisions are influenced by many factors, the most important of which are preference factors and geographic factors. Consider only these two most critical factors we divide a person’s travel decision-making process into three steps as follows:

  • 1.

    Preference

Experiments

In this section we will evaluate the effectiveness of our approach. The experiments are set up as the following.

Conclusions

In this paper, we design a novel and effective next new POI recommendation system by simulating user travel decision-making process. Our system considers two important factors that influence people’s choice of travel destination: preference factors and geographic factors, and integrates them into a unified recommendation process.

There are still some directions worth exploring and improving in the future. For example, how to better utilize heterogeneous information to construct a more accurate

Acknowledgments

This work is supported by the National Nature Science Foundation of China (61702368, 61170174), Major Research Project of National Nature Science Foundation of China (91646117), Natural Science Foundation of Tianjin, China (17JCYBJC15200, 15JCYBJC46500), Tianjin Science and Technology Major Projects and Engineering, China (15ZXZNCX00050), and the Opening Foundation of Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology, China (TJUT-KLICNST-K20170001). Finally, I

Xu Jiao received his BS and MS degrees in the School of Computer Science and Engineering, Tianjin University of Technology, Tianjin, P.R. China, in 2003 and 2010, respectively. He worked as an engineer at Tianjin Foreign Studies University, Tianjin, P.R. China, from 2003 to 2010. He was a lecturer in the Faculty of Fundamental Courses, Tianjin Foreign Studies University, Tianjin, P.R. China, from 2010 to 2015. He is currently a Ph.D. candidate in the School of Computer Science and Engineering,

References (39)

  • ZhangJ.D. et al.

    Core: Exploiting the personalized influence of two-dimensional geographic coordinates for location recommendations

    Inform. Sci.

    (2015)
  • SiY. et al.

    CTF-Ara: An adaptive method for POI recommendation based on check-in and temporal features

    Knowl.-Based Syst.

    (2017)
  • YeM. et al.

    Exploiting geographical influence for collaborative point-of-interest recommendation

  • ChengC. et al.

    Fused matrix factorization with geographical and social influence in location-based social networks

  • YuY. et al.

    A ranking based Poisson matrix factorization model for point-of-interest recommendation

    J. Comput. Res. Dev.

    (2016)
  • ZhangJ.D. et al.

    Igslr: personalized geo-social location recommendation:a kernel density estimation approach

  • ChoE. et al.

    Friendship and mobility:user movement in location-based social networks

  • YuanQ. et al.

    Time-aware point-of-interest recommendation

  • GaoH. et al.

    Exploring temporal effects for location recommendation on location-based social networks

  • ZhangJ.D. et al.

    Ticrec: A probabilistic framework to utilize temporal influence correlations for time-aware location recommendations

    IEEE Trans. Serv. Comput.

    (2017)
  • DebnathM. et al.

    Preference-aware poi recommen-dation with temporal and spatial influence

  • OzsoyM.G. et al.

    Time preference aware dynamic recommendation enhanced with location, social network and temporal information

  • YuanZ. et al.

    Location recommendation algorithm based on temporal and geographical similarity in location-based social networks

  • ShenglinZ. et al.

    STELLAR: Spatial-temporal latent ranking for successive point-of-interest recommendation

  • ZhangJ.D. et al.

    LORE: exploiting sequential influence for location recommendations

  • ZhangJ.D. et al.

    Spatiotemporal sequential influence modeling for location recommendations: A gravity-based approach

    ACM Trans. Intell. Syst. Technol.

    (2015)
  • YeM. et al.

    Location recommendation for location-based social networks

  • ZhangJ.D. et al.

    Geosoca: Exploiting geographical, social and Categorical correlations for point-of-interest recommendations

  • CuiL. et al.

    Periodic and successive point-of-interest recommendation under dual social group influences with matrix factorization

    Rev. Fac. Ing.

    (2016)
  • Cited by (37)

    • Pervasive computing of adaptable recommendation system for head-up display in smart transportation

      2022, Computers and Electrical Engineering
      Citation Excerpt :

      The prediction is made by a deep neural network that gives a more accurate result, and it obtains the user and features of items as the input and predicts the rating for evidence pointing. A simulated user travel decision-making process was discussed in [24] based on recommendation systems on POI (point of interest). This work is based on preference or geographic factors to employ the target user associated with the travel distance and probability.

    • Categorisation of cultural tourism attractions by tourist preference using location-based social network data: The case of Central, Hong Kong

      2022, Tourism Management
      Citation Excerpt :

      A table listing the details of the attractions is attached in the ‘Supplementary Material’. There are three main methods for social media data acquisition: (1) using published datasets (Jiao et al., 2019; Naserian et al., 2021), (2) data retrieval via application programming interfaces (APIs) (Chen et al., 2019; Shi et al., 2020), and (3) developing a web crawler (Carta et al., 2020). Compared with other platforms, such as Facebook and Twitter, Instagram shows unique demographic features with a predominant composition of younger generations (Karimkhani et al., 2014).

    • Network embedding enhanced intelligent recommendation for online social networks

      2021, Future Generation Computer Systems
      Citation Excerpt :

      This not only ensures the accuracy of the recommendation, but also reduces the computational time complexity. However, in the traditional recommendation systems, most methods only use a single network topology information and ignore other auxiliary information (such as user text information) [6,7]. For example, in real life, new users start to join social networks and there is no connection relationship with other users.

    • A POI group recommendation method in location-based social networks based on user influence

      2021, Expert Systems with Applications
      Citation Excerpt :

      Most existing researches have improved POI recommendation in LBSN by understanding and modeling the geographical influence, temporal effect, content information of POIs (Weiqing, 2017), and personality (Elahi et al., 2013). The most essential factor that distinguishes POI recommender systems from traditional RSs is geographical influence (Jiao, 2019). Users’ check-in histories contain valuable information exhibiting their preferences by analyzing their patterns (Zhu et al., 2018).

    • Probabilistic Time Geographic Modeling Method Considering POI Semantics

      2024, ISPRS International Journal of Geo-Information
    View all citing articles on Scopus

    Xu Jiao received his BS and MS degrees in the School of Computer Science and Engineering, Tianjin University of Technology, Tianjin, P.R. China, in 2003 and 2010, respectively. He worked as an engineer at Tianjin Foreign Studies University, Tianjin, P.R. China, from 2003 to 2010. He was a lecturer in the Faculty of Fundamental Courses, Tianjin Foreign Studies University, Tianjin, P.R. China, from 2010 to 2015. He is currently a Ph.D. candidate in the School of Computer Science and Engineering, Tianjin University of Technology, Tianjin, P.R. China. His research interest is personalized recommendation system.

    Yingyuan Xiao received the Ph.D. degree in computer science from Huazhong University of Science and Technology, P.R. China, in 2005. He is currently a professor in the School of Computer Science and Engineering, Tianjin University of Technology, P.R. China. He was a visiting scholar in the School of Computing at the National University of Singapore from 2009 to 2010. His research interests include advanced databases, personalized recommender systems, real-time systems, and mobile computing. He has published more than 100 journal and conference papers in these areas, including IEEE Transactions on Computers, IEEE Transactions on Parallel and Distributed Systems, Information Processing Letters, Journal of Classification, Soft Computing, Personal and Ubiquitous Computing, etc. He has served as a program chair of WAIM Workshop (International Workshop on Location-based Query Processing in Mobile Environments), a publicity chair of FSKD2009, and served as program committee member for a number of international conferences, including APWeb2011, APSCC2011, IEEE CloudCom2012 and ISI2013.

    Wenguang Zheng received his BS degree from University of Electronic Science and Technology of China, Chengdu, Sichuan, PR China, in 2010, and MS degree in National University of Singapore in 2011. He received the Ph.D. degree from the University of New South Wales, Sydney, Australia, in 2016. He is currently an associate professor in the school of computer science and engineering at Tianjin University of Technology, Tianjin, PR China. His research interests include computer architecture, embedded system, and machine learning.

    Hongya Wang received his BS and MS degrees in Electrical Engineering from Central China Normal University, Wuhan, Hubei, PR China, in 1998 and 2001, respectively, and the Ph.D. degree in Computer Science from Huazhong University of Science and Technology, Wuhan, Hubei, PR China, in 2005. He is currently a full professor in the School of Computer Science and Technology at Donghua University, Shanghai, PR China. He has published more than 30 papers in top journals and conferences such as TPDS, TC and CIKM. His research interests include data management, probabilistic algorithms for big data and machine learning.

    Ching-Hsien Hsu is a Professor in the department of computer science and information engineering at National Chung Cheng University, Taiwan; His research includes high performance computing, cloud computing, parallel and distributed systems, big data analytics, ubiquitous/pervasive computing and intelligence. He has published 200 papers in top journals such as IEEE TPDS, IEEE TSC, ACM TOMM, IEEE TCC, IEEE TETC, IEEE System, IEEE Network, top conference proceedings, and book chapters in these areas. He has been acting as an author/co-author or an editor/co-editor of 10 books from Elsevier, Springer, IGI Global, World Scientific and McGraw-Hill. Dr. Hsu is a Fellow of the Institution of Engineering and Technology.

    View full text