=Paper=
{{Paper
|id=Vol-2855/main_short_2
|storemode=property
|title=TripRec – A Recommender System for Planning Composite City Trips Based on Travel Mobility Analysis
|pdfUrl=https://ceur-ws.org/Vol-2855/main_short_2.pdf
|volume=Vol-2855
|authors=Rinita Roy,Linus W. Dietz
|dblpUrl=https://dblp.org/rec/conf/wsdm/RoyD21
}}
==TripRec – A Recommender System for Planning Composite City Trips Based on Travel Mobility Analysis==
https://ceur-ws.org/Vol-2855/main_short_2.pdf
ACM WSDM WebTour 2021, March 12th, 2021 Jerusalem, Israel 8
TripRec – A Recommender System for Planning Composite City
Trips Based on Travel Mobility Analysis
Rinita Roy Linus W. Dietz
rinita.roy@tum.de linus.dietz@tum.de
Technical University of Munich Technical University of Munich
Garching, Germany Garching, Germany
ABSTRACT destinations for suggesting trips to her. However, this is not person-
Location-based social networks (LBSNs) are rich sources of studying alised for users without prior Gmail accounts. Due to the inherent
travel mobility of people. With more users sharing updates about complexities, there is no application in the market that uses data-
their activities in LBSNs, there is a high availability of data to learn driven approaches for determining personalised, composite city
about their travel mobility patterns. This can help to improve travel trips for all users, motivating us to start working further in that
recommender systems, as we get a realistic impression of travelers’ direction. To tackle this challenge, we discover travel mobility pat-
travel behavior. We propose a system that recommends personalised terns from LBSNs and utilise them for computing personalised
city trips to different users by employing data-driven approaches. composite city trips.
Our web-based system recommends composite trips of 138 cities Traditionally, destination recommendation was subdivided into
all around the world. The application elicits user information and recommending regions [22], cities [7], point of interests (POIs) [1, 2,
preferences like home region, destination region, traveller type, 12], activities [18] or events [15]. Recommending POIs can mean rec-
maximum travel duration and fondness for different types of venues ommendation of next POI [12], top-k POIs using two common types
in a city, as inputs. Satisfying the user preferences and constraints, of recommender systems, viz., CF-based [1] and CBF-based [2], or
a suitable trip including an ordered list of cities with duration of composite POIs. The conversational DRS called CityRec developed
stay at each is determined, to be recommended to the user. by Dietz et al. [7] recommended only individual cities to users. The
RS developed by us recommends composite trips, specifically, multi-
KEYWORDS ple cities to be visited in order along with a recommended duration
of stay. A composite trip consists of a sequence of travel destina-
Destination Recommender Systems, City Trips, Data Mining, Per- tions. Composite tourist recommender systems (CTRSs) deal with
sonalisation choosing a number of travel destinations, selecting the sequence of
visit, and determining suitable duration of stay in each of the desti-
1 INTRODUCTION AND RELATED WORK nation. Researchers in the past developed CTRSs recommending
Destination recommender systems (DRSs) can help travellers to multiple countries [11, 22], or POIs [23, 24].
discover destinations to travel to. Depending on the type of data POI recommendations by Yu and Chang [24] were delivered
utilised, a recommendation model can be collaborative filtering to the users time-to-time, whereas, those provided by Wörndl et
(CF) or content-based filtering (CBF). The former model is typically al. [23] were displayed together at a time. CTRSs for POIs can rec-
based on explicit or implicit user feedback. CBF-based recommender ommend composite POIs with [3, 20] or without [23] constructing
systems (RSs) use the characteristic features of the items and the timed paths. Those with the timed paths suggest the time of arrival
preferences of a user before generating the recommendations for and time to leave the POI along with the sequence of POIs, determin-
them. The design of RSs, which earlier relied only on intuition- ing the duration of stay at each POI. CTRSs are mostly otherwise
based models, is now employing more data-driven approaches [4]. designed by solving tourist trip design problem (TTDP) [9, 10]. The
The latter involves analysis of large sets of data, interpreting and CTRS designed by us also uses the approach of solving TTDP with
incorporating them for building better decision-making strategies. the maximum travel duration constraint to recommend one or more
City tourism, also known as urban tourism involves travelling to cities to tourists.
the urban cities of different countries. It facilitates the development
of the cities to attract tourists. Moreover, with more than half of
the world population staying in urban areas [16], city tourism is
2 SYSTEM OVERVIEW
important for the economy as it brings employment to numerous We discuss the system in three broad stages – the data engineer-
individuals. On the other side, this is mainly interesting for tourists ing & analysis (pre-processing), CBF-based recommendation (peri-
who prefer to visit locations including architectures & monuments, processing), and user-centric evaluation of the web application
pubs & bars, restaurants & cafés etc. However, it is difficult for (post-processing).
people to determine desirable top destination cities to be visited for
their next trip. City RSs become useful in this context.
Google Trips1 collects data from Gmail account of a user and 2.1 Data Engineering, Analysis &
combines it with other features like crowd-sourced reviews about Pre-processing
Initially, we collect, analyse and clean the data to make it ready to
1 https://www.google.com/travel/ be used for the CBF-based recommendation.
Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
ACM WSDM WebTour 2021, March 12th, 2021 Jerusalem, Israel 9
Roy and Dietz
2.1.1 Mapping of World Regions. We divide six continents of the within the trip, 𝐹𝑏 𝑗 denotes value of the feature 𝐹 for block
world (except Antarctica) to 10 global regions, viz., North Amer- 𝑏 𝑗 , and 𝐹 designates 𝐴𝐸, 𝐹 𝐷, 𝑁 𝐿, 𝑂𝑅, or 𝐶𝐼 .
ica, South America, North Europe, Southwest Europe, Southeast After this characterisation, some of these trips are removed based
Europe, North Africa, South Africa, West Asia, East Asia and Ocea- on their poor qualities to assure a better quality of the dataset.
nia. Figure 1 displays these regions on the world map.
2.1.3 Identification of Regional Traveller Types. Unlike the previous
papers for clustering travellers [6, 8], in this paper, we segregate
trips by travellers from different home regions before identifying
the travel mobility patterns. We discover 10 prototype clusters for
the types of travellers around the world, after characterising the
trips followed by them. Next, we cluster the trip subsets and thus the
travellers using k-means clustering into suitable number of groups
chosen using silhouette index. This is followed by the identification
of the traveller types found in different regions. The 47 traveller
types so obtained from different regions of the world act as the
possible options for traveller types to be chosen from by a user of
Figure 1: World map annotated with our customised world our final RS application. The methodologies, traveller types and
regions their analysis can further be found in the elaborated discussion in
the master thesis by Roy [17].
2.1.2 Datasets – Cities & Trips. We consider 138 cities to be recom- 2.1.4 Calculation of Duration of Stays. The number of days to stay
mended to travellers of different types. The cities are attributed with at any city to be recommended to different types of travellers are
different features. Those involving the frequencies of venues, with pre-calculated and stored. At first, we compute the mean duration
different types of touristic values located in the cities, are called of stay at a city considering the trips by all the travellers of the
arts & entertainment (AE), food (FD), nightlife (NL), and outdoors same type having their home location in the same region. We do
& recreation (OR). The types of these venues are based on four of this for all the cities, for the different traveller types belonging to
the Foursquare venue categories2 . We divide each of the frequency each of the 10 regions. Altogether, we obtain 47 different values for
values by the total venue counts of all four types considered in the duration of stay at each city depending on the 47 traveller types.
respective cities. This is done to avoid bias due to the varied range In the trips we consider, not all traveller types visit all the cities.
of venue counts in different cities and check the prevalence of the However, we can find visits to all of the 138 cities in our database,
different types in each of the cities. Finally, for each city, stemming if we consider the travellers of all types. We do not intend to omit
from a number derived from Numbeo3 , the cost index (CI) values the possibility of recommendation of any of the 138 cities to any
are normalized between 0 and 100. traveller type. We update the duration of stay at a city with the
Trips are identified out of the check-ins from Twitter using a average stay by the travellers of all types belonging to the particular
data-mining approach [5, 21]. Each trip is annotated with different home region, if it is found to be zero by the current traveller type. If
characteristic features: it is still zero, we update the duration again with the mean duration
of stay in the city by the travellers of all types from all the home
• Mobility-based features – the features that help in analysing
regions. This is done for every city, for the 47 traveller types.
mobility patterns of travelling in the respective trips. This in-
cludes travel duration, displacement, radius of gyration, cities
2.2 Pre-processing & Content-based
visited, and countries visited [6].
• Traveller characteristics – the features that include informa- Recommendation
tion about the travellers of the identified trips. This includes This section explains the user inputs, CBF-based recommendation
home region of a traveller and the home ratio, providing the algorithm and the overview of the web application.
ratio of the number of check-ins at her home country to that
2.2.1 User Inputs. A user needs to input her preferences based on
at a location outside the home country.
which a CBF-based recommendation is provided to her. Following
• City-based features – features signifying the kind of places
are the inputs required for our algorithm:
visited during the trip. Since there can be multiple cities in
a trip, we calculate the average value 𝐹𝑖 for each city-based (1) Home region of the user, chosen from the 10 world regions.
feature within a trip 𝑖 using Equation 1. (2) Traveller type of the user that suits her the best, chosen from
the list of traveller types for those having their home same
Í𝑛 as the user’s home region.
𝑗=1 𝑛𝑏 𝑗 ∗ 𝐹𝑏 𝑗 (3) Destination region for the user’s desired trip, chosen again
𝐹𝑖 = Í𝑛 , (1)
𝑗=1 𝑛𝑏 𝑗 from the 10 world regions.
where 𝑛 is the distinct number of blocks (cities) within the (4) Maximum duration for the desired trip.
trip, 𝑛𝑏 𝑗 denotes the number of times block 𝑏 𝑗 is visited (5) The preference levels for the different city-based features.
2 https://developer.foursquare.com/docs/resources/categories 2.2.2 Recommendation Algorithm. After calculating the duration
3 https://www.numbeo.com/cost-of-living/ of stay at different cities for different traveller types, and after
Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
ACM WSDM WebTour 2021, March 12th, 2021 Jerusalem, Israel 10
TripRec – A Recommender System for Planning Composite City Trips Based on Travel Mobility Analysis
having the inputs from one user, we utilise the following steps as
part of the recommendation algorithm to plan a composite city trip
for her:
(1) Filtering Cities According to Destination Region – From
the 47 lists of cities with different duration of stays for dif-
ferent traveller types, we select the list based on the current
user’s home region and her travelling type. From that list
of 138 cities, we remove the ones which do not belong to
the region chosen by the user as her destination region. As
a result, we are left with a subset of cities considered further
for recommendation.
(2) Assigning Scores to Cities – For each city in the filtered
subset, we find the Euclidean distance between the city-based
feature vector (𝐶 = [𝐶𝐼𝑐𝑖𝑡 𝑦 , 𝐴𝐸𝑐𝑖𝑡 𝑦 , 𝐹𝑐𝑖𝑡 𝑦 , 𝑁𝑐𝑖𝑡 𝑦 , 𝑂𝑅𝑐𝑖𝑡 𝑦 ]) and
the user preference vector (𝑃 = [𝑃𝐶𝐼 , 𝑃𝐴𝐸 , 𝑃𝐹 , 𝑃 𝑁 , 𝑃𝑂𝑅 ]).
This is followed by using equation 2 to assign a score to
each of them. The simple score metric used here serves the
purpose of giving higher values to the cities whose feature
values are closer to the preferences of a user.
1
𝑠𝑐𝑜𝑟𝑒𝑐𝑖𝑡 𝑦 ← (2) Figure 2: Interaction between user and system
𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑐𝑖𝑡 𝑦 + 1
(3) Selection of Cities – The filtered cities are sorted based
on the scores assigned to them. Then the greedy selection visit some cities in North Europe within 16 days as Eurotrotters [17],
of the highly scored cities is done until the total duration who are travellers from Europe, travelling to many nearby cities
of stay at the selected cities does not exceed the maximum and countries.
travel duration input of the user. This constitutes the initial
list of selected cities. Some of the cities, that are far away
from others in the list and have only a single day as the
duration of stay for the current user, are removed from the
list. After the removal, if the constraints permit, more cities
are considered to be added to form the final list of selected
cities.
(4) Ordering of Cities – For ordering the selected list of cities,
we initially find different orders starting from each city in
the list as source and moving to the nearest one next. Then
we calculate the total distances to be covered for visiting the
cities in the different orders, and pick up the specific order
with the shortest distance to be covered. Figure 3: Exemplary recommendation by TripRec
2.2.3 TripRec Web Application. We develop TripRec , a data-driven
prototype web application to recommend composite city trips to The agreement questions in the feedback form follow the ResQue
different types of travellers using the discussed recommendation Questionnaire [14], a validated evaluation tool for RS:
strategy. A user interface (UI) facilitates the interaction between a
(Q1) The individual travel destinations recommended to me matched
user and the system. Figure 2 shows the interaction flows between
my interests
a user and the system through the UI for TripRec.
(Q2) The composite travel destinations recommended to me matched
While prompting a user to provide different inputs for eliciting
my interests
her preferences, the system also guides her with helpful information
(Q3) The recommended duration of stays at each city seems appro-
in every step while she uses the application. As a trip recommen-
priate for me
dation, the UI displays the cities, its corresponding countries, and
(Q4) I understood why the travel destinations were recommended
the duration of stay at each city in the order returned by the rec-
to me
ommendation algorithm. This is also accompanied by presenting
(Q5) I found it easy to tell the system what my preferences are
the order of visits to the recommended cities using Google Maps4 .
(Q6) TripRec allows me to modify my taste profile
If no recommendation is found for the specified inputs, the user is
(Q7) The layout and labels of the recommender interface are clear
asked to modify them and try again. Figure 3 shows an exemplary
(Q8) Overall, I am satisfied with this recommender system
recommendation result to a user from Southwest Europe, opting to
(Q9) I would use this recommender system again, when looking
4 https://www.google.com/maps for travel destinations
Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
ACM WSDM WebTour 2021, March 12th, 2021 Jerusalem, Israel 11
Roy and Dietz
The user needs to specify their level of agreement to the different (2) Users dissatisfied with the recommended duration of stays
statements on a five-point likert scale [13]. The responses check the (𝑄3) were comparatively more than those dissatisfied with
RS on the basis of quality of the recommended items, transparency, the recommended cities (𝑄1, 𝑄2).
ease of preference elicitation and revision, interface adequacy and (3) Maximum number of users have strongly agreed to having
attitudes of the user. There are also personal questions about the clear layouts and labels for the interface (𝑄7), followed by
age and gender of the user and a place to add additional comments. those strongly agreeing to being able to specify their pref-
erences to the system (𝑄5) and then modify them (𝑄6) as
2.3 User-centric Evaluation of Web Application well.
(4) A lot of users have agreed to have overall liked TripRec (𝑄8).
We conducted a user study for TripRec to examine the behaviours of
However, comparatively lesser people agreed to use the sys-
its users, their opinions about the system, and some characteristics
tem again (𝑄9) in real-life. People were more neutral about
of the services provided to them. Within a span of two weeks,
the latter, possibly because of the system being a research
we accumulated 217 recommendation requests from 113 unique
prototype that can recommend from just 138 cities.
users, 75 of whom provided feedback used for the evaluation of the
application. Finally, we determine how long the users interact with the system
in terms of interaction time and feedback time. We consider only
2.3.1 Different Users and their Behaviours. The application re- the final interaction time by each user. Eliminating an outlier record
ceived the maximum number of requests by people from West with interaction time of about 10 hours, we plot the histogram for
Asia, followed by those from Southwest Europe and Southeast Eu- interaction time as shown in Figure 4:Left, the mean interaction time
rope, whereas there were no users from Oceania. Other than West being 4 minutes and 40 seconds with 6 minutes standard deviation.
Asia, users wanted to go to the European regions, viz., Southwest Figure 4:Right shows the histogram for feedback time, the average
Europe, Southeast Europe, and North Europe the most. Irrespective being 5 minutes and 45 seconds with standard deviation 12 minutes
of home regions, the users usually tend to visit cities with low to and 25 seconds.
medium cost index. A lot of the users chose to follow the traveller
type vacationers [17], who are travellers making a short trip not so
far, possibly within their own home regions.
2.3.2 Analysing Recommendations Based on User Data. We analyse
the recommendations provided to users by TripRec based on their
preferences. We determine which cities are recommended the most
by calculating the recommendation ratio (RR) of the cities within
each destination region. RR of a city is the number of times it is
recommended divided by the total number of recommendations
within the corresponding destination region. We can see more Figure 4: Left: Interaction time histogram. Right: Feedback
variation of cities in the recommendation results when there are time histogram
more cities under a destination region in our database. Using our
user study data, we also find out, on an average, what proportion
of the total travel duration is the recommended duration of stay at 2.3.4 Qualitative Feedback Analysis. Some users gave additional
each city, for travellers from different regions. The recommended comments expressing their concerns or providing suggestions to
average duration of stay at a city divided by the average duration of improve our system. Few notable ones are summarised below with
a trip is called as the mean proportionate duration of stay (MPDS) at our remarks on top of that:
a city. Results show that the addition of more cities in the database (1) Exclude the countries already visited by a user from the
with diverse duration of stays can balance the MPDS at different recommendation list provided to her – this was out of scope,
cities. but can be considered later.
(2) Round sliders for providing preferences for city-based fea-
2.3.3 Quantitative Feedback Analysis. We find out how satisfied tures were difficult to handle in some mobile devices – our
the users were with our system based on their age groups. We prototype system was not yet designed for the specific needs
noticed that most of the users belonged to the age range between of all types of devices, but can later be made more responsive.
21 and 30 years, and the users aged over 40 years tended to get (3) Nearby cities should be recommended – the prototype data-
more satisfied with the system. base had only 138 cities to check the functionalities of the
Next, we compare the users’ agreement levels to the various feed- system. With more cities added, the system is supposed to
back questions. Users seem to have mostly agreed to the provided perform better.
questions in favour of TripRec. However, looking at the responses More information about TripRec and its user-centric evaluation
closely, we note the following points: can be found in the master thesis by Roy [17].
(1) Most of the users were satisfied with the individual recom-
mendations (𝑄1). However, the number of users satisfied 3 CONCLUSION & FUTURE WORK
with the composite recommendations (𝑄2) was compara- In this paper, we designed and developed the first destination rec-
tively lower. ommender system for computing personalised, composite city trips
Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
ACM WSDM WebTour 2021, March 12th, 2021 Jerusalem, Israel 12
TripRec – A Recommender System for Planning Composite City Trips Based on Travel Mobility Analysis
for any user, after analysing mobility data from location based social Characterization. In ACM RecSys Workshop on Recommenders in Tourism (Copen-
networks. The overall complexity of composite destination recom- hagen, Denmark) (RecTour 2019). 17–21.
[8] Linus W. Dietz, Avradip Sen, Rinita Roy, and Wolfgang Wörndl. 2020. Mining trips
mendations is very high, which is reflected in our system, which from location-based social networks for clustering travelers and destinations.
employs various data engineering steps, including characterisation Information Technology & Tourism (2020). https://doi.org/10.1007/s40558-020-
00170-6
of cities and trips, mapping of the different cities to 10 world regions [9] Damianos Gavalas, Charalampos Konstantopoulos, Konstantinos Mastakas, and
and identification of regional traveller types. We presented a novel Grammati E. Pantziou. 2014. A survey on algorithmic approaches for solving
algorithm for content-based composite city trip recommendations, tourist trip design problems. J. Heuristics 20, 3 (2014), 291–328. https://doi.org/
10.1007/s10732-014-9242-5
which we deployed in prototype web application that served as an [10] Daniel Herzog, Linus W. Dietz, and Wolfgang Wörndl. 2019. Tourist Trip Rec-
user-centric evaluation platform. ommendations – Foundations, State of the Art and Challenges. In Personalized
For the upcoming versions of TripRec, we can incorporate fea- Human-Computer Interaction, Miriam Augstein, Eelco Herder, and Wörndl Wolf-
gang (Eds.). de Gruyter Oldenbourg, Berlin, Germany, 159–182.
tures like the budget information of the recommended trip and the [11] Daniel Herzog and Wolfgang Wörndl. 2014. A Travel Recommender System for
flight booking options, which were out of scope for our present Combining Multiple Travel Regions to a Composite Trip. In Proceedings of the
1st Workshop on New Trends in Content-based Recommender Systems co-located
research and the prototype application. POIs specific to user pref- with the 8th ACM Conference on Recommender Systems, CBRecSys@RecSys 2014,
erences should be shown to the users, when demanded, for each Foster City, Silicon Valley, California, USA, October 6, 2014. 42–48.
of the recommended cities. This might enhance user satisfaction [12] David Massimo and Francesco Ricci. 2019. Clustering Users’ POIs Visit Tra-
jectories for Next-POI Recommendation. In Information and Communication
and ensure recommendation transparency. Furthermore, we plan to Technologies in Tourism 2019, Juho Pesonen and Julia Neidhardt (Eds.). Springer
develop strategies for computing recommendations when the user International Publishing, Cham, 3–14.
has specified a set of preferences that would currently return an [13] Victor R. Preedy and Ronald R. Watson. 2010. 5-Point Likert Scale. In Handbook of
Disease Burdens and Quality of Life Measures (New York, NY). Springer New York,
empty set. Such empty recommendations could happen due to the New York, NY, USA, 4288–4288. https://doi.org/10.1007/978-0-387-78665-0_6363
limited availability of only 138 cities for our prototype application. [14] Pearl Pu, Li Chen, and Rong Hu. 2011. A User-centric Evaluation Framework for
Recommender Systems. In Fifth ACM Conference on Recommender Systems (RecSys
We can add more cities to our database to provide more realistic ’11). ACM, New York, NY, USA, 157–164. https://doi.org/10.1145/2043932.2043962
recommendations. The world divisions could then also be made [15] Daniele Quercia, Neal Lathia, Francesco Calabrese, Giusy Di Lorenzo, and Jon
more granular using a touristic region model [19], given that each Crowcroft. 2010. Recommending Social Events from Mobile Phone Location
Data. In Proceedings of the 2010 IEEE International Conference on Data Mining
region would have more cities. Moreover, with more options of (ICDM ’10). IEEE Computer Society, Washington, DC, USA, 971–976. https:
cities in the database, we can take account of the distances between //doi.org/10.1109/ICDM.2010.152
the cities from the beginning during their selection. Further im- [16] Hannah Ritchie. 2018. Urbanization. Our World in Data (2018).
https://ourworldindata.org/urbanization.
provement can be in terms of the duration of stays recommended [17] Rinita Roy. 2020. A Recommender System for Planning Composite City Trips Based
to different users at the selected cities, making it more personalised. on Travel Mobility Analysis. Masterarbeit. Technische Universität München,
85748 Garching b. München.
Finally, we can also make our next prototype more responsive [18] Rinita Roy and Linus W. Dietz. 2019. Modeling Physiological Conditions for
for more devices other than web. More options can be provided to Proactive Tourist Recommendations. In Proceedings of the 23rd International
the users to elicit their interests, priorities, and posteriorities, for Workshop on Personalization and Recommendation on the Web and Beyond (Hof,
Germany) (ABIS ’19). ACM, New York, NY, USA, 25–27. https://doi.org/10.1145/
greater user satisfaction. 3345002.3349289
[19] Avradip Sen and Linus W. Dietz. 2019. Identifying Travel Regions Using Location-
Based Social Network Check-in Data. Frontiers in Big Data 2, 12 (June 2019).
https://doi.org/10.3389/fdata.2019.00012
REFERENCES [20] Von-Wun Soo and Shu-Hau Liang. 2001. Recommending a Trip Plan by Ne-
[1] Jie Bao, Yu Zheng, and Mohamed F. Mokbel. 2012. Location-based and Preference- gotiation with a Software Travel Agent. In Cooperative Information Agents V,
aware Recommendation Using Sparse Geo-social Networking Data. In Proceedings Matthias Klusch and Franco Zambonelli (Eds.). Springer Berlin Heidelberg, Berlin,
of the 20th International Conference on Advances in Geographic Information Systems Heidelberg, 32–37.
(Redondo Beach, California) (SIGSPATIAL ’12). ACM, New York, NY, USA, 199– [21] Lukas Vorwerk and Linus W. Dietz. 2021. An Interactive Dashboard for Traveler
208. https://doi.org/10.1145/2424321.2424348 Mobility Analysis. In ACM WSDM Workshop on Web Tourism.
[2] Liangliang Cao, Jiebo Luo, Andrew Gallagher, Xin Jin, Jiawei Han, and Thomas S [22] Wolfgang Wörndl. 2017. A Web-based Application for Recommending Travel
Huang. 2010. A worldwide tourism recommendation system based on geotagged Regions. In Adjunct Publication of the 25th Conference on User Modeling, Adapta-
web photos. In 2010 IEEE International Conference on Acoustics, Speech, and Signal tion and Personalization (Bratislava, Slovakia) (UMAP ’17). ACM, New York, NY,
Processing, ICASSP 2010 - Proceedings (ICASSP, IEEE International Conference USA, 105–106. https://doi.org/10.1145/3099023.3099031
on Acoustics, Speech and Signal Processing - Proceedings). 2274–2277. https: [23] Wolfgang Wörndl, Alexander Hefele, and Daniel Herzog. 2017. Recommending
//doi.org/10.1109/ICASSP.2010.5495905 a sequence of interesting places for tourist trips. J. of IT & Tourism 17, 1 (2017),
[3] Munmun De Choudhury, Moran Feldman, Sihem Amer-Yahia, Nadav Golbandi, 31–54. https://doi.org/10.1007/s40558-017-0076-5
Ronny Lempel, and Cong Yu. 2010. Automatic Construction of Travel Itineraries [24] Chien-Chih Yu and Hsiao-ping Chang. 2009. Personalized Location-Based Rec-
Using Social Breadcrumbs. In Proceedings of the 21st ACM Conference on Hypertext ommendation Services for Tour Planning in Mobile Tourism Applications. In
and Hypermedia (Toronto, Ontario, Canada) (HT ’10). Association for Computing E-Commerce and Web Technologies, Tommaso Di Noia and Francesco Buccafurri
Machinery, New York, NY, USA, 35–44. https://doi.org/10.1145/1810617.1810626 (Eds.). Springer Berlin Heidelberg, Berlin, Heidelberg, 38–49.
[4] Linus W. Dietz. 2018. Data-Driven Destination Recommender Systems. In Pro-
ceedings of the 26th Conference on User Modeling, Adaptation and Personalization
(Singapore, Singapore) (UMAP ’18). Association for Computing Machinery, New
York, NY, USA, 257–260. https://doi.org/10.1145/3209219.3213591
[5] Linus W. Dietz, Daniel Herzog, and Wolfgang Wörndl. 2018. Deriving Tourist
Mobility Patterns from Check-in Data. In Proceedings of the WSDM 2018 Workshop
on Learning from User Interactions. Los Angeles, CA, USA.
[6] Linus W. Dietz, Rinita Roy, and Wolfgang Wörndl. 2019. Characterisation of
Traveller Types Using Check-In Data from Location-Based Social Networks.
In Information and Communication Technologies in Tourism 2019, ENTER 2019,
Proceedings of the International Conference in Nicosia, Cyprus, January 30-February
1, 2019. 15–26. https://doi.org/10.1007/978-3-030-05940-8_2
[7] Linus W. Dietz, Myftija Saadi, and Wolfgang Wörndl. 2019. Designing a Con-
versational Travel Recommender System Based on Data-Driven Destination
Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).