Tourism Integrated Recommender System:

Setubal Peninsula Case Study

Mohamma Julashokri

, Suzana Monteiro Leonardi

and Pedro Seabra

Viatecla SA, Estr. Algazarra 72, 2810-013 Almada, Portugal

Faculdade de Letras, Universidade de Lisboa, Lisboa, Portugal

Keywords: Tourism, Recommender System, Collaborative Filtering, Content-based Filtering, Point of Interest.

Abstract: The diversity and a huge number of different places and attractions can make the decision for tourists difficult.

The recommender systems are developed to facilitate people's decision-making process that can help in the

area of tourism as well. In this paper, we proposed and implement a recommender system that works

integrated with a Setubal peninsula portal to help tourists to choose experiences and points of interest to visit.

The implementation was done using collaborative and content-based filtering to make recommendations

based on user profiles and activities within the portal.

1 INTRODUCTION

The tourism market has grown enormously during the

past 60 years. The number of international tourist

trips in 1950 was 25 million, however, this number in

2000 became 674 million, and increased to 1.235

million in 2016. From a profit perspective

international tourism in 1950 has a market of $ 2

billion and in 2000 became $ 495 billion but the

market increased to $1.220 billion in 2016. The

prediction for market growth in international travel is

3.3% per year until 2030 and will reach $1.8 billion.

(Huang et al 2017).

Recommender systems can help tourists to select

between a large number of destinations. In the tourism

market, the recommender systems suggest different

things like experiences, places to visit, hotels, etc.

using different methods like content-based filtering,

collaborative filtering, and knowledge-based filtering

or their composition (Santos et al 2016).

In the content-based filtering method,

characteristics of users' interests are extracted and

other items that have the same or similar

characteristics will be recommended (Park & Chang,

2009, Julashokri et al, 2011); While collaborative

filtering considers similarities between user profiles.

A user’s profile in collaborative filtering consists of

https://orcid.org/0000-0003-0012-9456

https://orcid.org/0000-0002-4434-5567

https://orcid.org/0000-0002-5771-3648

user-specified ratings to items and user demographics

and the recommendations will build based on

similarities between user profiles (Montaner et al.,

2003; Park & Chang, 2009, Julashokri et al, 2011).

Knowledge-based filtering is to use any kind of

knowledge in the recommendation-making

processes, like when the system has knowledge that

how a specific item relates to a specific user (Alpekin

& Buyukozkan, 2011).

The rest of this work is organized as follows.

Section 2 outlines the background and reviews related

works on customer profiles and recommendation

methods. Section 3 illustrates the proposed methods.

Section 4 describes the implementation of the

proposed model. Finally, Section 5 draws conclusions

and future works.

2 LITERATURE REVIEW

Many studies have dealt with recommender systems

in the tourism area. Some of them proposed mobile

recommender systems to help tourists to choose

points of interest and others focused on the

recommendation-making process.

Noguera et al (2012) present a novel mobile

recommender system that brings together a hybrid

Julashokri, M., Leonardi, S. and Seabra, P.

Tourism Integrated Recommender System: Setubal Peninsula Case Study.

DOI: 10.5220/0011102100003179

In Proceedings of the 24th International Conference on Enterprise Information Systems (ICEIS 2022) - Volume 1, pages 623-627

ISBN: 978-989-758-569-2; ISSN: 2184-4992

623

recommendation engine and a mobile 3D GIS

architecture. Their system gives users a 3D map-

based interface based on user location.

Nilashi et al (2017) proposed a recommendation

method based on multi-criteria collaborative filtering

to enhance the predictive accuracy of recommender

systems in the tourism domain using clustering,

dimensionality reduction, and prediction methods.

They also used a cluster ensembles approach to

improve the recommendation accuracy. Based on the

TripAdvisor dataset, they evaluated their model to

confirm that cluster ensembles can provide better

predictive accuracy.

Smirnov et al (2014) proposed a context-aware

recommender system that gathers information about

the point of interest from internet sources like

Wikipedia, wikivoyage, wikitravel, panoramio and

flickrm. In their system, tourists can rate attractions

that they like, or dislike and they used this

information to find similar users using the

collaborative filtering method. Smirnov et al (2014)

presented a mobile app that recommends point of

interest and the appropriate transportation to that

POIs using users’ context like location and weather

as well as user profiles like trip length, interaction

mode, etc.

Jorro-Aragoneses et al (2017) also proposed a

context-aware recommender system to recommend

leisure activities in Madrid, called Madrid Live. Their

system asks for restrictions and preferences of the

tourist and tries to satisfy restrictions to recommend a

set of activities based on tourist preferences. The

contextual information that the Madrid Live system

uses to make recommendations are time, location,

weather, users’ budget, and if they wish to use public

transport.

In Khallouki et al (2018), authors presented a new

approach to make context-aware mobile

recommender system that combines Internet of

Things (IoT) technologies with semantic web

services to predict the suitable recommends for user.

Alrasheed et al (2020) proposed a multi-level

tourism recommender system framework that

provides the user with a set of destinations liked by

similar users to allow constructing a list of potential

destinations. Then their system ranks the selected

destinations depending on user preference and user

constraint using data gathered from different Web

portals to consider the dynamic context of each trip.

3 PROPOSED MODEL

This paper proposes the creation of an integrated

recommendation system based on users’ profiles and

ratings to suggest both POIs and experiences based

on the accuracy level of that content (POI,

Experience, etc.). This algorithm shows the

recommendation building process in the current

work:

1: Choose the class of recommendations (can be

more than a class)

1.1: Content-based: if the customer booked

experience on that class or if he read an

article about that class

1.2: Collaborative-based: make an SVM model

based on customer profile (based on user

demographics, user bookings and user logs)

and predict the class of experience that this

user likes

2: Get the location of person and make

recommendation list based on (region, on make

itinerary)

3: Filter the recommendations if they are open at the

time, if the customer already visit that experience

Figure 1: High-level architecture of proposed model.

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

624

In Figure 1, a high-level architecture is shown.

A. Portal

The “Peninsula de Setubal” is implemented to share

the experience, articles, etc. to make people familiar

with Setubal attractions.

For receiving data from this portal to build a data

mining model in python or to different parts of the

recommender system in C#.Net, a list of APIs was

implemented.

The portal is responsible for calling the

recommender system for each user and then sending

the recommendations to the user using interaction

channels like email.

B. Data Mining Model

To implement the data mining model for predicting

customer preferences, the python application is

implemented. The application receives a data file

containing the user profiles labeled with the category

of the content (experience, route, article, …). This

historical data which is collected from user data in the

Portal uses for building an SVM model using the

scikit-learn library of python. Then an API that has

been implemented with Flask library will listen to

receive user data to call the model and prepare the

recommendation categories(classes) based on

customer profile.

C. Recommender System

The main part of the recommender system is an

application written in C#.Net.

This application has an API that listens to receive,

requests containing a user profile and will respond to

the request with a list of recommendations that should

send to the user.

In the first step, the application prepares the

categories of recommendations from two sources:

collaborative filtering and content-based filtering.

Collaborative filtering is implemented in python

and described before as python block. The content-

based filtering will give the list of categories that

users booked or have logged on to them.

After having the categories from two resources a

unique list with different weights for each list will be

prepared.

In the second step, the list of recommendations in

the categories will be extracted.

And in the third step, the recommendation list will

be filtered and the extra recommendations like the

ones that have already been booked by the user will

be removed.

4 MODEL IMPLEMENTATION

The proposed model is implemented to integrate

with the portal of “Enjoy Peninsula de Setubal”

which is a complete tourist guide to promote

attractions of Setubal peninsula in Portugal

(https://visitsetubalpeninsula.scriptorserver.com/).

After preparing the data using Portal API and

saving them in CSV file to start building the model,

first, we compared the different types of data mining

algorithms to see which one has the better answer on

our data.

We used the python scikit-learn library to build

our model using different algorithms:

 'LR': Logistic Regression

 'LDA': Linear Discriminant Analysis

 'KNN': K Nearest Neighbors Classifier

 'CART': Decision Tree Classifier

 'NB': Gaussian NB

 'SVM': Support Vector Machine

In the separated script we compare mentioned

algorithms on our data to see which one can have

better output results for us and we used k-Fold Cross-

Validation.

Cross-validation is a resampling procedure to

evaluate machine learning models based on a limited

data sample. Cross-validation is primarily used in

applied machine learning to estimate how the model

is expected to perform in general when used to make

predictions on data not used during the training of the

model.

Figure 2 shows the comparison between different

algorithms on sample data.

Figure 2: Comparison between different algorithms on

sample data.

Based on figure 2 the best results appeared with

LDA and SVM algorithms. As our data still test data

Tourism Integrated Recommender System: Setubal Peninsula Case Study

625

Figure 3: Python data mining model listen to receive new data sample for prediction.

and will be grown during the productional use of

portal we decided to use the SVM algorithm for now,

because the LDA algorithm supposes that the data

points have the same covariance, and the probability

density is normally distributed but SVM has no such

assumption.

Finally, we made our data mining model based on

SVM, and to have better performance on run time, we

build the model one time then use an API to receive a

new data sample and respond to the prediction for that

data sample. As is shown in Figure 3, we developed

the API using the Python Flask library.

5 CONCLUSION AND FUTURE

WORKS

Nowadays, Recommender systems play a major role

in any business that tries to sell products or services

to customers and plays in a huge and diverse market.

In the tourism area, the companies face with a huge

number of peoples that have totally different wishes

as well as loads of places and experiences which can

be suggested to these customers.

There are a huge number of destinations that can

make customers choose his-her next destination for

holidays.

The aim of this study is to propose and implement

a complete and integrated system that support all data

and process from booking of experience, reading of

the articles about POI, etc. to track the user feedback

and recommend new POIs and experiences based on

the user profile.

This system is making the recommendations

using data mining algorithms for collaborative

filtering and content-based filtering and works as part

of a complete and integrated system to support all

tourism needs to make the decision. The automated

DevOps test showed that the recommendation system

works correctly at the Unit and integration levels.

This integrated system will also have the location

track and will improve its recommendations based on

the user location and the time of the day in the future.

Also, after having more interactions with real

customers the datamining algorithm parameters will

be tuned; Because depending on data behavior the

input fields' impact in the data mining model should

change.

ACKNOWLEDGMENT

The research leading to these results has received

funding from the Portugal 2020 program, project

XPLORE_MARKET (ALT20-03-0247-FEDER-

024296).

REFERENCES

Alptekin, G.I., Buyukozkan, G., 2011. An integrated case-

based reasoning and MCDM system for Web based

tourism destination planning. Expert Systems With

Applications 38, 2125–2132.

Alrasheed, H., Alzeer, A., Alhowimel, A., Shameri, N.,

Althyabi, A. (2020). A Multi-Level Tourism

Destination Recommender System, Procedia Computer

Science, 170(333-340), https://doi.org/10.1016/j.procs.

2020.03.047

Berger H., Denk M., Dittenbach M., Merkl D., Pesenhofer

A. (2007) Quo Vadis Homo Turisticus? Towards a

Picture-based Tourist Profiler. In: Sigala M., Mich L.,

Murphy J. (eds) Information and Communication

Technologies in Tourism 2007. Springer, Vienna.

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

626

C.D. Huang, J. Goo, K. Nam, C.W. Yoo, Information &

Management 54, 757 (2017), smart Tourism: Traveler,

Business, and Organizational Perspectives.

Cufoglu, A. (2014). User Profiling-A Short Review. 108(3),

1–9.

Cufoglu, A., Lohi, M., & Everiss, C. (2017). Feature

weighted clustering for user profiling. International

Journal of Modeling, Simulation, and Scientific

Computing, 08(04), 1750056. https://doi.org/10.1142/

s1793962317500568

Cufoglu, A., Lohi, M., & Madani, K. (2009). A comparative

study of selected classifiers with classification accuracy

in user profiling. 2009 WRI World Congress on

Computer Science and Information Engineering, CSIE

2009, 3, 708–712. https://doi.org/10.1109/CSIE.2009.

954

Inoue, C. R. (2015). Tipos de revisão de literatura. Tipos de

Revisão de Literatura, 9. Retrieved from

http://www.fca.unesp.br/Home/Biblioteca/tipos-de-

evisao-de-literatura.pdf

Jorro-Aragoneses, J.L., Díaz Agudo, M.B., Recio García,

J.A. (2017). Madrid live: a context-aware recommender

systems of leisure plans, 29th International Conference

on Tools with Artificial Intelligence (ICTAI), Boston,

DOI: 10.1109/ICTAI.2017.00125

Julashokri, M., Fathian, M., Gholamian, M.R., Mehrbod, A.

(2011). Improving electronic customers' profile in

recommender systems using data mining techniques,

Management Science Letters 1, 449–456

Khallouki, H., Abatal, A., Bahaj, M. (2018). An Ontology-

based Context awareness for Smart Tourism

Recommendation System, Proceedings of the

International Conference on Learning and Optimization

Algorithms: Theory and Applications(LOPAL '18),

43(1-5), https://doi.org/10.1145/3230905.3230935

Kotsiantis, S. B. (2007) Supervised Machine Learning: A

Review of Classification Techniques. Informatica 31,

249-268

K.R.Ananthapadmanaban, A., & Srivatsa, S. K. (2011).

Personalisation of User Profile: Creating User Profile

Ontology for Tamilnadu Tourism. International

Journal of Computer Applications, 23(8), 42–47.

https://doi.org/10.5120/2903-3808

Linden, Greg & Smith, Brent & York, Jeremy. (2003).

Amazon. com recommendations: Item-to-item

collaborative filtering. Internet Computing, IEEE. 7.

76-80.

Martín-Bautista, M. J., Kraft, D. H., Vila, M. A., Chen, J.,

& Cruz, J. (2003). User profiles and fuzzy logic for web

retrieval issues. Soft Computing - A Fusion of

Foundations, Methodologies and Applications, 6(5),

365–372. https://doi.org/10.1007/s00500-002-0190-x

Nilashi, M., Bagherifard, K., Rahmani, M., Rafe, V. (2017).

A recommender system for tourism industry using

cluster ensemble and prediction machine learning

techniques, Computers & Industrial Engineering, 109

(357-368), https://doi.org/10.1016/j.cie.2017.05.016

Montaner, M., Lopez, B., & De La Rosa, J. L. (2003). A

taxonomy of recommender agents on the Internet.

Artificial Intelligence Review, 19(4), 285-330.

Noguera, J.M., Barranco, M.J., Segura, R.J., Martínez, L.

(2012). A mobile 3D-GIS hybrid recommender system

for tourism, Information Sciences, 215(37-52),

https://doi.org/10.1016/j.ins.2012.05.010

Norma Sandra Ferreira de Almeida. (2002). As pesquisas

denominadas estado de arte. Educação & Sociedade,

79(257–272), 257–272. Retrieved from

http://www.scielo.br/pdf/es/v23n79/10857.pdf

Park, Y.-J., & Chang, K.-N. (2009). Individual and group

behavior-based customer profile model for

personalized product recommendation. Expert Systems

with Applications, 36(2), 1932 -1939.

Ricci, Francesco. (2002). Travel recommender systems.

IEEE Intelligent Systems. 17. 55-57.

Santos, F., Almeida, A., Martins, C., Oliveira, P. (2016).

Gonçalves, R., Tourism Recommendation System

based in user’s profile and functionality levels, C3S2E

'16: 9th International Conference on Computer Science

& Software Engineering, 93-97, https://doi.org/

10.1145/2948992.2948995

Schiaffino, S., Scientific, N., & Scientific, N. (2009).

Artificial Intelligence An International Perspective.

5640(January 2009). https://doi.org/10.1007/978-3-

642-03226-4

Schiaffino S., Amandi A. (2009) Intelligent User Profiling.

In: Bramer M. (eds) Artificial Intelligence An

International Perspective. Lecture Notes in Computer

Science, vol 5640. Springer, Berlin, Heidelberg

Smirnov, A., Kashevnik, A., Ponomarev, A., Teslya, N.,

Shchekotov, M., Balandin, S.I.(2014). Smart space-

based tourist recommendation system: Application for

mobile devices, Internet of things, smart spaces, and

next generation networks and systems, Springer

International Publishing, Cham, 40-51, DOI:

10.1007/978-3-319-10353-2_4

Tourism Integrated Recommender System: Setubal Peninsula Case Study

627