A Survey of Point-of-Interest Recommendation
in Location-Based Social Networks
Yonghong Yu
Tongda College
Nanjing University of Posts and Telecommunications
yuyh@njupt.edu.cn
Xingguo Chen
School of Computer Science and Technology
School of Software
Nanjing University of Posts and Telecommunications
chenxg@njupt.edu.cn
Abstract
With the rapid development of mobile devices, global
position system (GPS) and Web 2.0 technologies,
location-based social networks (LBSNs) have attracted
millions of users to share rich information, such as
experiences and tips. Point-of-Interest (POI) recom-
mender system plays an important role in LBSNs since
it can help users explore attractive locations as well as
help social network service providers design location-
aware advertisements for Point-of-Interest. In this pa-
per, we present a brief survey over the task of Point-of-
Interest recommendation in LBSNs and discuss some
research directions for Point-of-Interest recommenda-
tion. We first describe the unique characteristics of
Point-of-Interest recommendation, which distinguish
Point-of-Interest recommendation approaches from tra-
ditional recommendation approaches. Then, according
to what type of additional information are integrated
with check-in data by POI recommendation algorithms,
we classify POI recommendation algorithms into four
categories: pure check-in data based POI recommenda-
tion approaches, geographical influence enhanced POI
recommendation approaches, social influence enhanced
POI recommendation approaches and temporal influ-
ence enhanced POI recommendation approaches. Fi-
nally, we discuss future research directions for Point-
of-Interest recommendation.
Introduction
With the rapid development of mobile devices, global po-
sition system (GPS) and Web 2.0 technologies, location-
based social networks (LBSNs) have become very popular
and attracted lots of attention from industry and academia.
Typical location-based social networks include Foursquare,
Gowalla, Facebook Place, and GeoLife, etc.. In LBSNs,
users can build connections with their friends, upload pho-
tos, and share their locations via check-in for points of inter-
est (e.g., restaurants, tourists spots, and stores, etc.). Besides
providing users with social interaction platforms, it is more
desired for LBSNs to make use of the rich information (so-
cial relationships, check-in history and so on) to mine users’
preferences on locations and recommend new places where
users may be interested in. The task of recommending new
Copyright
c
2015, Association for the Advancement of Artificial
Intelligence (www.aaai.org). All rights reserved.
interesting places is referred as point-of-interest (POI) rec-
ommendation. POI recommender systems have played an
important role in LBSNs since they can not only meet users’
personalized preferences for visiting new places, but also
help LBSNs to increase revenues by providing users with
intelligent location services, such as location-aware adver-
tisements.
Although recommender systems have been widely studied
and successfully adopted by many e-commerce web sites,
such as Amazon, Netflix, Last.fm and Taobao etc., POI rec-
ommender systems have just emerged recently. Differing
from traditional recommender systems, POI recommender
systems have the following unique characteristics.
• Geographical Influence. As the Tobler’s First Law of
Geography reported that ”Everything is related to every-
thing else, but near things are more related than distant
things” (Tobler 1970). For LBSNs, the Tobler’s First Law
of Geography implies that users prefer to visit nearby lo-
cations rather than distant ones and users may be inter-
ested in POIs surrounded a POI that users prefer. Geo-
graphical Influence is the most important characteristic
that distinguish POI recommender systems from tradi-
tional recommender systems and heavily effect users’ vis-
iting behaviors.
• Frequency Data and Sparsity. In traditional recom-
mender systems, user generally expressed their prefer-
ences by explicitly providing ratings for items (e.g., book,
movie, music and so on), which are converted to user-
item rating matrix. The ratings are often numerical values
and fall into a numerical range, such as [1,5]. The higher
rating corresponds the better satisfactory. Unlike to tradi-
tional recommender systems, a user’s preferences are re-
flected by the frequency of check-in for locations, which
are often transformed to user-location check-in frequency
matrix. The frequency data have a large range compared
with ratings. For example, user may check in thousands
of times at some locations, while user may check in few
times for other locations. In addition, the sparsity of user-
location check in frequency matrix is dramatically higher
than that of user-item rating matrix, which leads to big-
ger challenge for POI recommendation. For example, the
sparsity of Netflix data set is around 99%, while the spar-
sity of Gowalla is about 2.08 × 10
−4
.
Trajectory-Based Behavior Analytics: Papers from the 2015 AAAI Workshop