=Paper=
{{Paper
|id=Vol-1405/paper-01
|storemode=property
|title=Location-Aware Recommendation Systems: Where We Are and Where We Recommend to Go
|pdfUrl=https://ceur-ws.org/Vol-1405/paper-01.pdf
|volume=Vol-1405
|dblpUrl=https://dblp.org/rec/conf/recsys/Rodriguez-Hernandez15
}}
==Location-Aware Recommendation Systems: Where We Are and Where We Recommend to Go==
https://ceur-ws.org/Vol-1405/paper-01.pdf
Location-Aware Recommendation Systems: Where We Are
and Where We Recommend to Go
María del Carmen Sergio Ilarri
Rodríguez-Hernández University of Zaragoza, Spain
University of Zaragoza, Spain silarri@unizar.es
692383@unizar.es
Raquel Trillo-Lado Ramón Hermoso
University of Zaragoza, Spain University of Zaragoza, Spain
raqueltl@unizar.es rhermoso@unizar.es
ABSTRACT 1. INTRODUCTION
Recommendation systems have been successfully used to The progressive development of mobile computing tech-
provide items of interest to the users (e.g., movies, music, nologies has allowed the emergence of Location-Based Ser-
books, news, images). However, traditional recommenda- vices (LBS). LBS attempt to provide useful and customized
tion systems do not take into account the location as a information in contexts where the location is an important
relevant factor when providing suggestions. On the other factor to bear in mind, such as scenarios related to health
hand, nowadays, there exist an increasing amount of geo- issues, working environments, entertainment, personal life,
referenced data and users are usually interested only in nearby and so on. The locations of moving objects are typically ob-
items (e.g., restaurants, museums, cinemas). Hence, the tained by using information obtained by the mobile devices
emergence of location-aware recommendation systems have through the communication network used for data transmis-
acquired a great attention by the research community in the sion or by exploiting geographical positioning systems (e.g.,
last decade. GPS sensors, beacon techniques).
In this paper, we provide a survey of location-aware rec- Recommendation Systems (RS) have been a main focus
ommendation systems in mobile computing scenarios. Firstly, of research, as these systems gradually reduce the existing
we describe briefly the fundamentals of recommendation sys- information overload (information available on the Internet,
tems. Then, we introduce some of the most relevant existing data provided by sensors of different types or other users,
approaches for location-aware recommendation. Moreover, etc.), by recommending to the users personalized items of
we present the main applications of this type of systems interest (e.g., movies, music, books, news, images) based on
in several recommendation scenarios, such as music, news, their preferences. With the advent of e-commerce, the com-
restaurants, etc. Finally, we discuss new avenues and open bination of recommendation system techniques and LBS has
issues in the area. been of significant interest for researchers. The inclusion of
the location dimension in these types of systems allows ob-
Categories and Subject Descriptors taining more effective recommendations, so bringing about
the emergence of a new field of research called Location-
H.3 [Information Storage and Retrieval]: Information Aware Recommendation Systems (LARS).
filtering In this paper, we provide a survey on location-aware rec-
ommendation systems for mobile computing. The rest of
Keywords the paper is organized as follows. Section 2 provides some
fundamentals about the technological context. In Section 3,
Location-aware recommendation systems, mobile comput-
we present an overview of related works. Then, we classify
ing, open issues
different approaches by application domain in Section 4. In
Section 5, we discuss future perspectives of LARS. Finally,
we present our conclusions in Section 6.
2. BACKGROUND
2.1 Traditional Recommendation Systems
Recommendation Systems (RS) are applications aimed at
suggesting items of interest to users (e.g., products, ser-
vices). Recommendations are considered an important sup-
port for users’ decision making (e.g., decide which products
Copyright held by the author(s).
to buy, which book to read next, which movie to watch) [17].
LocalRec’15, September 19, 2015, Vienna, Austria. They are important from both the business perspective and
from the user’s perspective, as they can boost purchases but can be continuously moving but also the items (e.g., if the
also alleviate the information overload experienced by the items are taxi cabs).
users. The location can be associated to the physical position of
Based on how recommendations are calculated, RS are the user when he/she rates an item (e.g., a book rated by a
generally classified into three well-known categories [2], as user from home), to the location of an item (e.g., the position
explained in the following. of a restaurant rated), or to both. The framework proposed
Collaborative filtering: the user is provided with items in [19] classifies location-based ratings in three categories:
consumed in the past by other users with similar tastes and
preferences (user-based collaborative filtering). Another pos- • Spatial ratings for non-spatial items. Represented by
sibility is to recommend items based on the similarity with the tuple (user, ulocation, rating, item), where uloca-
other items that the user has liked in the past (item-based tion is the user’s location.
collaborative filtering); this similarity is computed by ana-
lyzing the ratings given to the items by the users. • Non-spatial ratings for spatial items. Stated by the tu-
Content-based recommendation: recommendations are based ple (user, rating, item, ilocation), where ilocation rep-
on the similarity between the searched item and other items resents an item’s location.
the user liked in the past. As opposed to the case of item-
• Spatial ratings for spatial items. Represented by the
based collaborative filtering, this item similarity is computed
tuple (user, ulocation, rating, item, ilocation). In this
by comparing the contents (features) of the items.
case, the location of the user and the location of the
Hybrid recommendation approaches: these methods com-
item are both relevant.
bine both collaborative and content-based algorithms, to
benefit from the advantages of each paradigm while trying
The users of LARS can receive implicit or explicit rec-
to avoid their specific disadvantages.
ommendations. On the one hand, implicit recommendations
Although major advances have been accomplished by us-
(push-based recommendations) are proactive recommenda-
ing, fine-tuning, and extending traditional recommendation
tions that the user receives without submitting explicit re-
techniques, they can fail when estimating the relevance of
quests to the system. On the other hand, explicit recom-
a certain item in some situations (e.g., where the users are
mendations (pull-based recommendations) are reactive rec-
interested only in nearby items). In particular, they run
ommendations, obtained as an answer to a query explic-
into severe problems when tackling scenarios with dynamic
itly submitted by the user (e.g., “I need a restaurant”). In
variables, such as the location of the user, time, weather, or
both cases, the set of recommendations provided to the user
other users’ opinions.
should be monitored and kept up-to-date, as the relevant
recommendations may change due to movements of the user
and/or target items.
2.2 Location-Aware Recommendation Systems Currently, several real-world recommendation systems use
To alleviate the problems of traditional RS mentioned the location as an important parameter for the suggestion
above, considerable efforts have been invested in the last of relevant items. Well-known examples are Google Now
years, creating a new research line called Context-Aware (http://www.google.com/landing/now/), Foursquare (http:
Recommendation Systems (CARS) [3]. These novel methods //foursquare.com/), and Yelp (http://www.yelp.com).
take into consideration the need of including the context of Finally, it is interesting to indicate that GPS trajectories
the user and/or the context of items in the process followed obtained from the user’s mobile logs can facilitate the dis-
to calculate accurate recommendations. Among the differ- covery of interesting patterns about the user [9, 32, 33], that
ent aspects that can be considered to represent the context may be further used to calculate recommendations.
of a recommendation process, the location of users and/or
items has been proved to be of special importance to suggest 3. DOMAIN-INDEPENDENT APPROACHES
relevant recommendations [16].
Location-Aware Recommendation Systems (LARS), illus- FOR LARS
trated in Figure 1, take into account the spatial properties In the recent years, thanks to advances of mobile devices,
(locations) of users and/or items to calculate proper recom- ubiquitous computing, and wireless communication tech-
mendations. The emergence of LARS comes from the fact nologies, a significant number of works have been carried
that users typically prefer nearby items (e.g., restaurants, out in the field of LARS. An example is the system pre-
museums, cinemas), as the effort needed to reach items close sented in [19, 29], which exploits location-based ratings to
to their physical positions will be smaller. Moreover, it may provide recommendations. To obtain spatial ratings, the au-
happen that only nearby items are relevant or that items thors applied an approach of user partitioning based on the
located far have a short spatio-temporal relevance. For ex- user locality, the scalability to large numbers of users, and
ample, a suggestion about a specific parking space provided the influence of the users, to control the size of the neigh-
to a driver searching for parking could become obsolete in borhood. For spatial items, a travel penalty was applied
a short time if the parking space is not nearby (while the (favoring the closest items). The collecting process of the
user drives towards the parking spot, it can be occupied by spatial ratings was motivated by the study carried out on
another vehicle). In general, LARS can be considered as an the MovieLens dataset (http://grouplens.org/datasets/
extension of traditional recommendation systems, and an movielens), that associates the locations with the user’s
important subset of CARS that focuses on the dimension ZIP codes (i.e., spatial ratings), and the Foursquare dataset
location in the multidimensional context. In LARS, the rat- (https://developer.foursquare.com/), which contains in-
ing is modeled as a function in terms of the item, user and formation about the places visited by users (i.e., spatial rat-
location f : U × I × L → R. Notice that not only the users ings for spatial items). Recently, and along the same vein,
Automatic data acquisition Generic architectures and User interfaces studies Security and privacy
and context exploitation middleware
User profiles
Sensor GPS
Data acquisition
Location-based
ZIP code 𝑈𝑈 × 𝐼𝐼 × 𝐿𝐿 → 𝑅𝑅
ratings
Mobile users
GPS User profiles
Sensors RFID
Evaluation
Trajectory GPS
diversity
LARS accuracy CHALLENGES
Item recommendations effectiveness
efficiency
usability
News Shopping POIs Music Taxi Car parking
Figure 1: Overview of LARS
the authors of [5] presented LARS*, that also recommends URLs) to try to address the handicap of popularity bias
items based on location-based ratings, by using user par- without affecting the performance. Moreover, recently, the
titioning and travel penalty techniques. In this case, the authors of [37] proposed a location-sensitive recommenda-
location is obtained from the IP address of the user’s mobile tion approach in ad-hoc social network environments.
device. With the development of the Web 2.0, some works fo-
A similar goal was pursued in [39], where the authors pre- cus on the combination of mobile technologies with tradi-
sented LA-LDA, a location-aware probabilistic generative tional social networks, giving rise to Location-Based So-
model that uses location-based ratings to model user pro- cial Networks (LBSN) [7], such as Foursquare, Facebook
files to produce recommendations (e.g., suggestions about Places (https://www.facebook.com/places/), and others.
restaurants) as well as to mitigate the well-known cold start The emerge of this new kind of social networks allows to
problem. They considered the three types of location-based connect with friends, share locations (and/or photos, videos,
ratings proposed in [19] (i.e., spatial user ratings for non- etc.), receive recommendations of places (e.g., restaurants),
spatial items, non-spatial user ratings for spatial items, and etc. The main research topic covered is how to effectively
spatial user ratings for spatial items). In [18], the authors combine the information provided by social networks to offer
proposed a location-based service recommendation model more accurate recommendations. For example, a user could
(LBSRM) that combines relevant elements of LBS and rec- trust particularly the recommendations offered by his/her
ommendation technologies. Firstly, the model filters infor- friends, but not all the user’s connections are necessarily
mation based on the user’s location, and then it recom- real friends. Analyzing in depth how information about the
mends relevant mobile information services by using clus- user’s social interactions in real-time (e.g., a tweet or photo
tering techniques. With a similar spirit, the authors of [13] published by the user, a conversation with a friend) could
recently integrated LBS with recommendation techniques to be exploited in the context of LARS is an issue that has not
present a hybrid recommendation model. been explored in depth so far.
Other approaches consider the impact of the locations not We conclude this section with some final examples. First,
only as a pre-filtering step but directly on the application of a Markov-based technique presented in [1] improves the qual-
collaborative filtering. For example, [11] uses Voronoi dia- ity of location-aware recommendation systems by using the
grams to decompose the user’s space and then it uses them location information of items. In the Markov model, the au-
in a spatially-aware collaborative filtering algorithm; specif- thors consider each item as a state. The states are defined
ically, they explored the concept of spatial autocorrelation as the history of items viewed (or visited) by the users, and
to cluster similar values on a map, by using statistical mea- the transition probability is calculated according to the pref-
sures. In this approach, the ZIP code of the area is used erences (likes) of items by the users in the past. In general,
to identify the user’s location. A location-aware collabora- the recommendation approach suggests the items with the
tive filtering was also proposed in [27], which uses the user’s highest likelihood estimation, by taking account the loca-
location to recommend web content in real-time, increasing tion (i.e., a greater geographical distance among the items
the diversity of recommendations; specifically, the authors decreases the probability estimation). In [23], a collabora-
determine the diversity using the Levenshtein edit distance tive filtering recommendation approach is presented, focus-
between attributes of items (e.g., locations, tags, titles and ing on the specific case of suggesting geospatial locations
(e.g., latitude and longitude) where mobile users can take tion and the transportation way (e.g., driving a car, riding
photos. The final list of locations to recommend must be a bicycle, or walking).
within a (user-defined) suitable distance from the physical A location-based and preference-aware recommendation
position of the user. Instead of exploiting the users’ loca- system that suggests venues (e.g., restaurants and shopping
tions, the authors used three million geotagged photos taken malls) within a geospatial range was presented in [6]. It
from smartphones (i.e., photos implicitly containing geoco- learns the user preferences automatically from the user’s lo-
ordinates). In [31], data mining techniques (e.g., clustering cation history and infers the user’s expertise (e.g., in cate-
models) were used to recommend items to the mobile users gories such as Chinese food and shopping mall) in several
by considering the user’s location. Finally, [24, 25] presented cities. During the recommendation process, the system fil-
an improvement of collaborative filtering that combines the ters the candidate local experts in a geospatial range (de-
user’s geographical information and the content of items in fined by the user) and suggests the venues that match the
order to learn location-based user group preferences, con- user’s preferences and the social opinions of the selected lo-
sidered by the authors as a rating distribution of a group of cal experts. This type of system has the advantage of pro-
items. According to the study performed with the Movie- viding venues not only near the area where users live, but
Lens dataset, the user group preference has strong correla- also in cities unknown to them. A similar goal was pur-
tion with the location of the user. sued in the Location-Content-Aware Recommendation Sys-
tem (LCARS) proposed in [40], which recommends venues
(e.g., restaurants) or events (e.g., concerts and exhibitions)
4. LARS IN SPECIFIC DOMAINS within the city of the query initiator, by using the probabil-
In this section, we discuss several domains where location- ity of influence of the personal interests and local preferences
aware recommendation systems have been applied. Firstly, of the users. One of the main goals is to alleviate the data
we consider the recommendation of generic POIs (Points sparsity problem (the new city problem) based on the loca-
of Interest) in Section 4.1. Then, we analyze in Section 4.2 tion and content information of spatial items.
relevant references for the tourism domain. Afterwards, Sec- Specifically focused on the restaurant domain, [15] pro-
tion 4.3 focuses on news recommendation, and we mention posed a location-based recommendation architecture for dy-
in Section 4.4 several approaches proposed in the literature namic and ubiquitous environments. The authors combine,
for the shopping domain. Finally, we present in Section 4.5 in the proposed architecture, the ideas of location, person-
works related to other domains. alization, and content-based recommendation. As a final
example, the PECITS system [35] provides location-aware
4.1 LARS for the Recommendation of POIs recommendations of POI paths (e.g., a list of several con-
nections that the user could take to reach a certain POI, by
One of the most common application domains of LARS is
using public transportation and by foot) in Bolzano (Italy).
suggesting interesting points (e.g., restaurants) around the
user. For instance, a collaborative location-aware filtering
approach to recommend POIs to mobile users was proposed 4.2 LARS for the Tourism Domain
in [16], which exploits the location as a relevant element In the tourism domain the recommendation process im-
for the recommendation of items (e.g., restaurants) near the plies suggesting a set of products or services that support
user’s current location. The approach proposed is the result traveling and tour planning (e.g., attractions, accommoda-
of combining user-based collaborative filtering techniques tions, restaurants, and activities). For example, the authors
with a location-based partitioning method (i.e., it allows of [20, 22] integrated tourism mobile commerce and location-
an adequate rating database partitioning based on the lo- aware features into a traditional recommendation system to
cation), with the goal of achieving a high scalability. That provide real-time recommendations for visitors, by taking
work validates the hypothesis that users who live nearby into account the locations and the ratings of the attrac-
tend to visit the same local places. The proposal in [10] at- tions. Similarly, an architecture for location-based recom-
tempts to solve the problem of location-based context-aware mendation was proposed in [41], which supports personal-
recommendations of POIs by using a multiagent system ar- ized tour planning for mobile tourism applications by us-
chitecture [36]; the use of agents facilitates the collection ing rule-based recommendation techniques. Along the same
of POIs’ information available on the Web. Another exam- line, the authors of [9] present a system that recommends
ple is the location-dependent collaborative filtering system touristic places based on the user’s visiting history in differ-
presented in [34], that analyzes the mobile user’s moving fea- ent regions (e.g., cities or countries). To recommend loca-
tures (e.g., moving direction, position, and speed, obtained tions, a set of geotags (manually set on a map or automat-
through a GPS receiver) and the POIs, in order to recom- ically obtained form the GPS device) representing the lati-
mend to the mobile user those items of interest that are in tude and longitude where a user took a photo is exploited.
a region near the user’s current position and in the same This is considered useful to plan a touristic visit to a new
direction. In the rest of this section, we mention some other city or country.
examples.
An ubiquitous location-based recommendation algorithm 4.3 LARS for the Recommendation of News
that suggests relevant places to mobile users is presented Most LARS use the user preferences and the distance be-
in [30]. The system, named “I’m feeling LoCo”, consid- tween the current user’s location and the positions of the
ers the user profile and the places near him/her during the items for the recommendation of relevant items. However,
recommendation process. It automatically infers the user’s it is not usual to enrich the previous approach by using ex-
preferences (by mining social network profiles) and considers isting relations between items and tagged locations (e.g.,
spatio-temporal constraints in the recommendation process. geographical metadata of news articles), which could have
The physical constraints are delimited by the user’s loca- an impact on the recommendations.
Thus, the authors of [4] proposed an interesting spatial weather (e.g., bad, fair, and good) as context parameters.
model for location-based serendipitous recommendation of The location of the user is determined by using GPS.
news articles. For that purpose, they studied the existing There exist also some attempts to use the location for
associations between the user’s current location and the lo- recommendation in e-learning environments. The approach
cation data available in the geographical metadata of the in [14] recommends educational materials and peer learners
news articles. The introduction of serendipity in traditional who are nearby, by using RFID to detect the learner’s en-
collaborative filtering implies modifying the recommenda- vironmental objects and his/her location. The system also
tion approach to discover the novelty (or the surprise) and allows the learners to share knowledge, interact, collaborate,
useful items for the user, by sacrificing accuracy. exchange individual experiences, and visualize the objects
A location-based social networking system for mobile de- that surround the learner, the space of learning resources,
vices, named Sindbad, was proposed also in the field of and the distance to possible peer helpers.
news [28]. With Sindbad, the user can receive friends’ news
based on their locations, as well as messages posted by his/her
friends. Moreover, its recommendation system also sug- 5. FUTURE PERSPECTIVES
gests spatial items (e.g., restaurants) and non-spatial items In the following, we discuss some perspectives of interest
(e.g., movies) based on the users’ locations, the items’ loca- that should attract further research in the near future (see
tions, and the ratings provided by friends. For that purpose, Table 1 for a summary).
the location-aware recommendation module LARS proposed
in [19] was used. Automatic data acquisition and context exploitation.
Overall, we believe that location-aware recommendation
4.4 LARS for Shopping Recommendation systems could be more effective if the characteristics of the
In the field of mobile commerce (m-commerce), several dynamic environment were effectively exploited. In a mobile
types of LARS have been designed and presented in the liter- environment, the location information of the items and/or
ature to suggest a variety of products and services that may users is dynamic, and therefore constantly changing. Hence,
be of interest to users. An example is the location-aware such information should be updated with a certain frequency,
recommendation system presented in [38], that recommends using external sources such as sensors (e.g., GPS, RFID).
vendors’ web pages to interested customers in mobile shop- However, the use of sensors to obtain the dynamic infor-
ping. Another example is CityVoyager [33], a recommenda- mation needed is not sufficiently exploited in some cases.
tion system based on the user’s location history, which is For example, [11, 16] consider the ZIP code to identify the
obtained by using a GPS device. It recommends shops to user’s location, which is a coarse-grain location. Further-
the users based on the locations of previous shops visited. more, most works related to LARS, despite using locations
In order to avoid the need to type text, along with the during the recommendation process, do not detail how they
associated spelling problems and possible ambiguity, when were acquired (e.g., see [13, 18, 27, 31]). The acquisition and
the user needs to specify the types of items he/she is in- automatic discovery of user preferences (which may change
terested in, an interesting proposal was presented in [42]. from one location to another) from several external data
Specifically, the location-based shopping recommendation sources (e.g., social networks, sensors), based on the use of
system proposed uses an image of the desired item (e.g., data mining techniques, is a major research challenge. Thus,
shoes, clothes) provided by the user, as the query, as well a process that automatically acquires a rich set of data would
as the smartphone’s GPS coordinates, to recommend retail allow improving the effectiveness of the recommendations, as
shops (with information including their GPS coordinates, well as alleviating the cold start problem.
promotions, and special offers) to mobile users. Finally, the quality of the recommendations could be fur-
ther improved by enriching the user profile with additional
context features besides the location dimension (e.g., the
4.5 LARS for Other Scenarios transport way, the weather, the time). The intuitive idea
Finally, it should be highlighted that, although the do- is that, by exploiting more information about the user pref-
mains examined in the previous subsections are the most erences in different contexts, the recommendations obtained
common ones, there are other possible use cases. For ex- can be more appropriate for the current user’s context. How-
ample, in the area of music, the authors of [8] tackled the ever, more research work is needed to explore this path. For
problem of providing location-dependent music recommen- example, the impact of having more or less context informa-
dations by using emotional tags related to the music and tion should be analyzed, and automatic methods are needed
the places of interest. With this idea, they developed a mo- to capture the context variables (e.g., we cannot expect that
bile location-aware recommendation system named Playing- the user will explicitly provide all his/her contextual infor-
Guide, that suggests and plays appropriate music for a place mation when rating an item).
of interest for the user (e.g., the user might hear a specific
music while visiting a place of interest in a city). Evaluation.
Another interesting work is Motivate [21], which presents Regarding evaluation, there are still significant research
a context-aware mobile recommendation system that pro- challenges to be addressed. Firstly, over time, RS have be-
motes a healthy lifestyle. It recommends different kinds of come more complex, by considering new parameters during
useful advices to the user (e.g., take a break, walk/cycle to the recommendation process, such as the location. In the
a park, go to a museum), by considering the location of the same way, the metrics for the evaluation of these systems
user, the activities in the user’s agenda (e.g., go to work, should also probably be more complex. However, researchers
work, have lunch, go home, have dinner, and busy), the continue using traditional measures (e.g., MAE, RMSE, pre-
time (e.g., the start and end time of an activity), and the cision, recall, and F1 score) to evaluate location-aware rec-
Challenges State of the art
1) Automatic data LARS could be more effective if the characteristics of the
acquisition and context dynamic environment were effectively captured and exploited.
exploitation: representation, Examples of related contributions:
acquisition, and enrichment of
data dynamically. -Exploiting GPS trajectories: [9, 32, 33]
-GPS sensing: [6, 8]
2) Evaluation: evaluation There is a need to use evaluation measures different from the
measures adjusted to dynamic classical ones, adjusted for the evaluation of LARS. Moreover,
environments, the datasets used for evaluation are usually still the same
context-enriched data sets. datasets used to evaluate traditional recommendation systems
(e.g., MovieLens and Foursquare). Examples of related
contributions:
-Diversity measure: [27]
-Usability questionnaire: [8, 18, 30]
-Continuous query processing performance: [19]
3) User interfaces: proper It is necessary to design suitable user interfaces (i.e., simple and
design of user interfaces for intuitive) for LARS, in order to avoid overloading the user with
mobile devices and dynamic information. Examples of related contributions:
environments.
-Usability evaluation of interfaces: [8, 30]
4) Security and privacy: The study and application of techniques to ensure location
ensuring the location privacy privacy and user security are important challenges to consider in
and user security. the development of LARS. Examples of related contributions:
-For recommendation systems in general: [26]
-No relevant work specific to LARS has been identified
5) Generic architectures Despite the efforts, there is still no implemented architecture
and middleware: emerge of that facilitates the development of LARS for mobile
generic architectures. environments. An adaptable architecture that could be extended
and customized for several application scenarios would be really
useful. Examples of related contributions:
-Proposal of a generic framework: [12].
Table 1: Summary of challenges related with LARS
ommendation systems. So, we believe that an interesting The fields of mobile computing and recommendation sys-
research direction could be the emergence of new evalua- tems have evolved in a quite independent way. However,
tion measures. For example, combining metrics, such as the when considering LARS, it is clear that traditional recom-
accuracy and the diversity with the latency, or including mendation techniques should be completed with other data
location parameters in existing measures, could be an in- management techniques applied in mobile computing. As an
teresting area to analyze. Moreover, most works focus on example, it should be noted that a location can refer to the
the evaluation of the effectiveness of the recommendations, current continuously-changing physical position of a user, an
but in mobile environments the usability and efficiency are item, or both. This is particularly relevant in typical mobile
also relevant aspects to evaluate: timely suggestions could environments, where the user and/or the item can be mov-
be more important than perfect suggestions but with a long ing [8, 30, 34]. For example, consider the case of a user who
delay. is walking down the street and uses a mobile application
Secondly, the datasets used for evaluation are usually the that suggests to him/her an appropriate taxi in real-time;
same datasets used to evaluate traditional recommendation in this case, both the user and the target items may be mov-
systems (e.g., MovieLens and Foursquare). Hence, it is nec- ing. As another example, if we consider applications such as
essary to generate new datasets containing location informa- the recommendation of parking spaces to drivers, estimat-
tion (related to items, users, or both) to evaluate LARS. The ing the spatio-temporal relevance of the parking spaces is a
problem aggravates if we consider the evaluation of CARS, key issue (parking spots released recently and close to the
which require datasets enriched with significant context in- location of the user should be preferred).
formation. Real datasets could be collected more easily by
a mobile recommendation system if the user’s context data User interfaces.
are automatically detected, as suggested in the previous re- From the perspective of mobile applications, user inter-
search challenge. Furthermore, the definition of realistic syn- faces designed for recommendation purposes (explicit or im-
thetic data generators, or even crowdsourcing data collection plicit recommendations) should be simple and easy to under-
through videogames (gamification), could be explored. stand. However, very few studies have evaluated the usabil-
ity of interfaces in the context of recommendations [8, 30], or
Bridging the gap between mobile computing and LARS. have studied in depth the best way to present the informa-
tion. Hence, we believe that this could be a relevant research
line to take into account during the design of location-aware
recommendation systems. For example, location-aware rec- [3] G. Adomavicius and A. Tuzhilin. Context-aware
ommendation systems are usually designed for mobile phone’s recommender systems. In F. Ricci, L. Rokach,
screens. So, an important element to consider is the need B. Shapira, and P. B. Kantor, editors, Recommender
to visualize only a few recommendations (not a long list of Systems Handbook, pages 217–253. Springer, 2011.
suggestions), to avoid overloading the user by crowding the [4] Y. A. Asikin and W. Wörndl. Stories Around You:
screen with information, but at the same time those recom- Location-based serendipitous recommendation of news
mendations should be representative and diverse. Similarly, articles. In 2nd International Workshop on News
another problem is how to allow the user to easily spec- Recommendation and Analytics (NRA), pages 1–8.
ify his/her needs regarding the type of items that he/she CEUR Workshop Proceedings, 2014.
requires (in pull-based recommendations), for example by [5] S. S. Avhad and S. R. Durugkar. Lars∗:
using a keyword-based search interface which correctly in- Location-aware recommendation system. International
terprets the user’s intention. Journal Of Engineering, Education And Technology,
3(2):1–6, 2015.
Generic architectures and middleware. [6] J. Bao, Y. Zheng, and M. F. Mokbel. Location-based
In this field, most works are location-aware recommenda- and preference-aware recommendation using sparse
tion approaches and prototype systems that focus on a spe- geo-social networking data. In 20th International
cific application domain (e.g., music, tourism, POIs, news, Conference on Advances in Geographic Information
shopping). Despite some efforts to generalize this, there is no Systems (SIGSPATIAL), pages 199–208. ACM, 2012.
implemented architecture that facilitates the development [7] J. Bao, Y. Zheng, D. Wilkie, and M. F. Mokbel. A
of location-aware recommendation systems for mobile envi- survey on recommendations in location-based social
ronments. We believe that this aspect should be analyzed, networks. GeoInformatica, 19(3):525–565, 2015.
given the interest of having a generic solution that can be [8] M. Braunhofer, M. Kaminskas, and F. Ricci.
extended and adapted to different application domains [12]. Location-aware music recommendation. International
The previous list does not intend to be exhaustive. For ex- Journal of Multimedia Information Retrieval,
ample, security and privacy is another hot topic of research 2(1):31–44, 2013.
which has not been extensively studied so far in the field [9] M. Clements, P. Serdyukov, A. P. de Vries, and
of LARS, even though the user’s location may need to be M. J. T. Reinders. Personalised travel
shared to retrieve suitable recommendations.
recommendation based on location co-occurrence.
CoRR, abs/1106.5213:1–30, 2011.
6. CONCLUSIONS [10] H. Costa, B. Furtado, D. Pires, L. Macedo, and
In this work, we have provided a survey of location-aware A. Cardoso. Context and intention-awareness in POIs
recommendation systems for mobile environments. We first recommender systems. In 6th ACM Conference on
described the basics of LARS and some generic approaches. Recommender Systems, 4th Workshop on
Then, we presented a number of location-aware recommen- Context-Aware Recommender Systems (RecSys), 2012.
dation systems for several scenarios. Finally, several future [11] J. Das, S. Majumder, and P. Gupta. Voronoi based
perspectives and challenges, that we believe should guide location aware collaborative filtering. In 3rd National
upcoming research steps, were discussed. Conference on Emerging Trends and Applications in
In the last decade, location-aware recommendation ap- Computer Science (NCETACS), pages 179–183. IEEE,
proaches made an important progress thanks to significant 2012.
efforts developed by the research community. Nevertheless, [12] M. del Carmen Rodrı́guez-Hernández and S. Ilarri.
more research is needed to solve existing difficulties and de- Towards a context-aware mobile recommendation
sign systems able to obtain more effective recommendations. architecture. In I. Awan, M. Younas, X. Franch, and
We hope that this survey will encourage further efforts. C. Quer, editors, Mobile Web Information Systems,
volume 8640 of Lecture Notes in Computer Science,
7. ACKNOWLEDGMENTS pages 56–70. Springer, 2014.
[13] R. Duan, R. S. M. Goh, F. Yang, Y. K. Tan, and
This work has been supported by the CICYT project
J. F. B. Valenzuela. Towards building and evaluating
TIN2013-46238-C4-4-R, DGA-FSE, and a Banco Santander
a personalized location-based recommender system. In
scholarship held by Marı́a del Carmen Rodrı́guez Hernández.
IEEE International Conference on Big Data (Big
Data), pages 43–48. IEEE, 2014.
8. REFERENCES [14] M. M. El-Bishouty, H. Ogata, and Y. Yano. Perkam:
[1] A. Abbasi, A. Javari, M. Jalili, and H. R. Rabiee. Personalized knowledge awareness map for computer
Enhancing precision of markov-based recommenders supported ubiquitous learning. Educational Technology
using location information. In International & Society, 10(3):122–134, 2007.
Conference on Advances in Computing, [15] A. Gupta and K. Singh. Location based personalized
Communications and Informatics (ICACCI), pages restaurant recommendation system for mobile
188–193. IEEE, 2014. environments. In International Conference on
[2] G. Adomavicius and A. Tuzhilin. Toward the next Advances in Computing, Communications and
generation of recommender systems: A survey of the Informatics (ICACCI), pages 507–511. IEEE, 2013.
state-of-the-art and possible extensions. IEEE [16] T. Horozov, N. Narasimhan, and V. Vasudevan. Using
Transactions on Knowledge and Data Engineering, location for personalized POI recommendations in
17(6):734–749, 2005. mobile environments. In International Symposium on
Applications and the Internet (SAINT), pages [30] N. S. Savage, M. Baranski, N. E. Chavez, and
124–129. IEEE, 2006. T. Höllerer. I’m Feeling LoCo: A location based
[17] P. B. Kantor, L. Rokach, F. Ricci, and B. Shapira. context aware recommendation system. In G. Gartner
Recommender Systems Handbook. Springer, New York, and F. Ortag, editors, Advances in Location-Based
USA, 2011. Services, Lecture Notes in Geoinformation and
[18] M.-H. Kuo, L.-C. Chen, and C.-W. Liang. Building Cartography, pages 37–54. Springer, 2012.
and evaluating a location-based service [31] N. Shabib and J. Krogstie. The use of data mining
recommendation system with a preference adjustment techniques in location-based recommender system. In
mechanism. Expert Systems with Applications, International Conference on Web Intelligence, Mining
36(2):3543–3554, 2009. and Semantics, pages 28:1–28:7. ACM, 2011.
[19] J. J. Levandoski, M. Sarwat, A. Eldawy, and M. F. [32] Y. Takeuchi and M. Sugimoto. An outdoor
Mokbel. LARS: A location-aware recommender recommendation system based on user location
system. In 28th International Conference on Data history. In 1st International Workshop on Personalized
Engineering (ICDE), pages 450–461. IEEE, 2012. Context Modeling and Management for UbiComp
[20] X. Li, Z. Mi, Z. Zhang, and J. Wu. A location-aware Applications (UbiPCMM), pages 91–100. CEUR
recommender system for tourism mobile commerce. In Workshop Proceedings, 2005.
2nd International Conference on Information Science [33] Y. Takeuchi and M. Sugimoto. CityVoyager: An
and Engineering (ICISE), pages 1709–1711. IEEE, outdoor recommendation system based on user
2010. location history. In J. Ma, H. Jin, L. T. Yang, and
[21] Y. Lin, J. Jessurun, B. de Vries, and H. Timmermans. J. J. P. Tsai, editors, Ubiquitous Intelligence and
Motivate: Towards context-aware recommendation Computing, volume 4159 of Lecture Notes in
mobile system for healthy living. In Fifth Computer Science, pages 625–636. Springer, 2006.
International Conference on Pervasive Computing [34] C.-C. Tuan, C.-F. Hung, and T.-C. Kuei. Location
Technologies for Healthcare (PervasiveHealth), pages dependent collaborative filtering recommendation
250–253. IEEE, 2011. system. In International Conference on Future
[22] J. M. Noguera, M. J. Barranco, R. Segura, and Network Technologies, Qingdao, China, 2011.
L. Martı́nez. A Location-Aware Tourism Recommender [35] G. Tumas and F. Ricci. Personalized mobile city
System Based on Mobile Devices, chapter 7, pages transport advisory system. In W. Höpken, U. Gretzel,
34–39. World Scientific, 2012. and R. Law, editors, Information and Communication
[23] T. Phan, J. Zhou, S. Chang, J. Hu, and J. Lee. Technologies in Tourism, pages 173–183. Springer,
Collaborative recommendation of photo-taking 2009.
geolocations. In 3rd ACM Multimedia Workshop on [36] G. Weiss. Multiagent Systems. MIT Press, Cambridge,
Geotagging and Its Applications in Multimedia MA, USA, 2013.
(GeoMM), pages 11–16. ACM, 2014. [37] L. Yang, F. Hao, S. Li, G. Min, H. Kim, and S. Yau.
[24] Z. Qiao, P. Zhang, Y. Cao, C. Zhou, and L. Guo. An efficient approach to generating location-sensitive
Improving collaborative recommendation via recommendations in ad-hoc social network
location-based user-item subgroup. Procedia Computer environments. IEEE Transactions on Services
Science, 29(0):400–409, 2014. Computing, 8(3):520–533, 2015.
[25] Z. Qiao, P. Zhang, J. He, Y. Cao, C. Zhou, and [38] W.-S. Yang, H.-C. Cheng, and J.-B. Dia. A
L. Guo. Combining geographical information of users location-aware recommender system for mobile
and content of items for accurate rating prediction. In shopping environments. Expert Systems with
23rd International Conference on World Wide Web Applications, 34(1):437–445, 2008.
Companion (WWW Companion), pages 361–362. [39] H. Yin, B. Cui, L. Chen, Z. Hu, and C. Zhang.
ACM, 2014. Modeling location-based user rating profiles for
[26] N. Ramakrishnan, B. J. Keller, B. J. Mirza, A. Y. personalized recommendation. ACM Transactions on
Grama, and G. Karypis. Privacy risks in recommender Knowledge Discovery from Data, 9(3):19:1–19:41,
systems. IEEE Internet Computing, 5(6):54–62, 2001. 2015.
[27] T. Sandholm and H. Ung. Real-time, location-aware [40] H. Yin, B. Cui, Y. Sun, Z. Hu, and L. Chen. LCARS:
collaborative filtering of web content. In Workshop on A spatial item recommender system. ACM
Context-awareness in Retrieval and Recommendation Transactions on Information Systems,
(CaRR), pages 14–18. ACM, 2011. 32(3):11:1–11:37, 2014.
[28] M. Sarwat, J. Bao, A. Eldawy, J. J. Levandoski, [41] C.-C. Yu and H. ping Chang. Personalized
A. Magdy, and M. F. Mokbel. Sindbad: A location-based recommendation services for tour
location-based social networking system. In ACM planning in mobile tourism applications. In T. D. Noia
SIGMOD International Conference on Management of and F. Buccafurri, editors, 10th International
Data, pages 649–652. ACM, 2012. Conference on E-Commerce and Web Technologies
[29] M. Sarwat, J. J. Levandoski, A. Eldawy, and M. F. (EC-Web), volume 5692 of Lecture Notes in Computer
Mokbel. LARS∗: An efficient and scalable Science, pages 38–49. Springer, 2009.
location-aware recommender system. IEEE [42] T. Zuva, O. O. Olugbara, S. O. Ojo, and S. M.
Transactions on Knowledge and Data Engineering, Ngwira. Image content in location-based shopping
26(6):1384–1399, 2014. recommender systems for mobile users. Advanced
Computing: An International Journal, 3(4):1–8, 2012.