Personalized City Tour Planner

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320751356

10 SIGNALS: A personalized city tours prototype

Conference Paper · August 2017


DOI: 10.1109/ColComCon.2017.8088207

CITATIONS READS

2 206

3 authors, including:

Laura Martinez García


Fundación Universitaria Tecnológico Comfenalco
8 PUBLICATIONS   12 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

10 SIGNALS: A sightseeing tours Recommender System View project

All content following this page was uploaded by Laura Martinez García on 16 July 2018.

The user has requested enhancement of the downloaded file.


10 SIGNALS: A Personalized City Tours Prototype
Laura Martinez Garcia, MsC Jhonatan Montes Serna, Student
Vicerectoria de Investigaciones Vicerectoria de Investigaciones
Fundación Universitaria Tecnológico Comfenalco, Fundación Universitaria Tecnológico Comfenalco,
Cartagena, Colombia Cartagena, Colombia
lmartinezg@tecnologicocomfenalco.edu.co montesj@tecnocomfenalco.edu.co

Valentina Codutti, Student


Vicerectoria de Investigaciones
Fundación Universitaria Tecnológico Comfenalco,
Cartagena, Colombia
valentinacoduttitheran@gmail.co

Abstract— A recommender system is presented using a of the person. The recommender system (10 SIGNALS) uses a
stereotyping model to create tourist classes in order to match up route planning to assist tourists to plan a trip in a city
users with other user profiles according to the gender, age and destination according to his/her preferences. This method
tourism preferences. The prototype permits suggest points of permits suggest points of interest and create a route planning
interest and create a route planning in a city destination when
when user is classified in a category.
users are classified in a category. This approach is implemented
as a tourist tool called 10signals which offers personalized city
tours in Cartagena, Colombia. In this work, our goal is to construct a technique to
personalized travel planning system through the stereotyping
Keywords—Recommender System, KNN, Point Of Interest, model (predefined social group approach) to obtain an initial
Tourist, Stereotyping. list of points of interest to travel in the destination city. The
stereotyped model presented is based on K-means, K-Nearest
I. INTRODUCTION Neighbor (KNN) algorithm and google maps API. As a case
Recommender systems are programs which attempt to study, we proposed a technological prototype that
recommend items (products or services) to users (individuals recommends POIs in the city of Cartagena.
or businesses) by predicting a user's interest in an item based
The paper is organized as follows: Section 2 shows the
on related information and interactions [1]. In the context of
state of the art; in section 3 and 4 presents the results of the
tourist, recommender system is useful to give information
proposal, finally, some conclusions and future research
about touristic places according to the user preferences.
directions are outlined.
Recommender systems in tourism can recommend points II. RELATED WORKS
of interest (POIs) and POIs route planning. Several systems
are focused on recommend points of interest, which through a This section briefly reviews related work in the areas
list of recommendations, evaluated POIs and then apply a directly associated to recommendation of touristic activities.
recommender process by tour groups [2]. On the other hand,
based route planning systems offer an automatic sequence of According to the author in [4] recommender systems for
points of interest to visit by tourists considering your user the tourism domain can be classified in two groups: systems
profile and historical data. that suggest single items like: Triplehop’s TripMatcher and
Vacation Coach [5] and systems that suggest a set of items:
The user information is stored in user profile, usually Expedia (www.expedia.com), DieToRecs [6], and Trip@dvice
gathered in three ways: it can be explicitly captured by asking [7]. The authors also proposed a recommender system allows
the user directly for; the system can try to associate the user the tourist to receive a group of itineraries made up of a set of
with a predefined social group that has well known correlated events (procession, festival, special market, etc.). In
preferences (stereotyping) or new information about the user City Trip Planner [8] the system suggests points of interest
can be obtained in an implicit way by observing his/her after capturing the tourist’s trip constraints and interests
interaction with the system [3]. This proposal is based on through a small questionnaire.
stereotyping technique to create profiles tourist in order to
general preferences categories according to the gender and age

978-1-5386-1060-2/17/$31.00 ©2017 IEEE


In literature, there are approaches to resolve the problem D: Woman tourist between 48 years and 58 years
of generating according to the users. In [9] a recommender E: Male tourist between 18 years and 25 years
system is focused on constraint satisfaction is e-Tourism [9], F: Male tourist between 26 years and 36 years
based on the generation of a personalized tourist plan for the G: Male tourist between 37 years and 47 years
city of Valencia in Spain. H: Male tourist between 48 years and 58 years

In [10], present a prototype that stores information (such 3. To match similar users to groups of tourists and
as duration, cost, and availability, dates and times) about provide recommendations, we create the user’s
travel products in a relational database. Other examples are: neighborhood, calculating the similarity between each
INTRIGUE [11] and Smart Tourist Agenda Recommender user. Recommendations are computed based on the
(STAR) [12], recommender systems that suggest itineraries stereotype and then normalized given the user affinity
around the city of Turin (in north Italy), and the tourist guide to a stereotype. The model in order to provide
in [13] that support users visiting the city of Salamanca personalized content to the user learns of him/her
(Spain). thought an algorithm based on KNN (K-Nearest
Neighbors).
Other relevant research projects with respect to tourist
itineraries have been reported: in [14], a method for deriving a 4. The data collection comes from various sources: for
single multi-modal tourist itinerary is proposed considering a POI's is a Google Maps API and datasets to classify
variety of constraints and following a genetic algorithm-based users.
approach. P-Tour [15], dynamic tour guide (DTG) [16] and
City Trip Planner [17] personalized user preferences about B. Algorithm
specific types of sites are not taken into account; however, In first instance, the algorithm calculates the neighbors
they derive routes day-by-day. The authors in [18], proposed a through kNN library for NodeJS. This method is called in a
heuristic solution for the orienteering problem, that is, they request POST that takes 2 arguments and returns an object
only consider a single tourist itinerary. with a type value and percentage of the accuracy. Then, the
algorithm compares a JSON Object of user profile with similar
Other group of system recommender used an ontological tastes using Euclidean distances, returning top K closest
approach to resolve the problem of generate tourism itineraries objects according to the stereotype categories.
are: SigTur/E-Destination, use domain ontology to guide the A RESTFUL service was implemented in NodeJS Express
recommendation process, which permits to make inferences which uses a library that implements the KNN. This service is
about the correspondence between the characteristics of an consumed using HTTP POST Request. The interface contains
activity and a certain user profile. SigTur/E-Destination makes the next parameters (Figure 1):
a knowledge-level analysis of the user preferences, including
processes that make bottom-up and top-down propagation of • The Messages are in JSON format.
the preferences over the concepts of the ontology. • The first input parameter is an object JSON with the
information of the user to classify.
III. TOURISM RECOMMENDER SYSTEM
• The second input parameter is an array of profiles.
A. Recommendation Approach
• The output is a JSON object (user stereotype).
Group recommendation techniques consider preferences of
the single-user, thus the group model [20]. In our case, the Once the neighbors are obtained, the system gathered a
system recommends tourist itineraries through the Stereotype user profiles and estimates an itinerary route with some Point
Model maps individual user into one of the predefined groups. of Interest (POIs) that user may be interested according to
The process to identify user profiles and then generate his/her stereotype.
recommendation is the follow:
Consequently, the best rated is performed, ordering the
1. The initial stereotypes are manually defined involving recommended POIs list by user preference in descending order
tourism domain expert identifying several variables in in a map. Finally, the output is a sequence of attractions or
tourist that represents relevant user information and spots to be visited, through a google map, filtered to the
preference elicitation. tourist's according user preferences (interest, day of visit, cost,
etc.).
2. The cluster analysis is focuses on grouping objects,
detecting the optimal number of groups and their
composition based on the similarity. In this case, the
results of the process identify 8 clusters (classes) that
can be divided the objects (Tourist):
A: Woman tourist between 18 years and 25 years
B: Woman tourist between 26 years and 36 years
C: Woman tourist between 37 years and 47 years
Fig. 2. Domain-Oriented reference architecture for generating tourist
itineraries

Fig. 1. Group recommender algorithm using KNN D. Prototype


The aim of the recommender system is to assist the user to
C. System Architecture planning a trip in a city destination according to the user
Domain-oriented reference architecture is using in the preferences. The main tasks of the system are: Pre-Trip
proposal according to the following components (figure 2): planning, generation of activities in the destination city,
generate customized automatic routes, be guided to Point of
• Domain: The domain contains information about the Interest in the map, get information, buy ticket to Point of
business domain. The entities and provider are held Interest, and rate Point of Interest and travel agency services.
here. Providers represent the services interfaces, in our The recommendation process starts when the user is
case the algorithms and suppliers sites. The entities categorized. Then the system predicts the points of interest
represent the functionality of recommending points of and the estimated time to make the visit. The system displays
interest through similarity algorithms; include routes the map image, showing the points of interest listed.
and similarities between users.
Once the list of recommended items is converted into a
• Infrastructure: Infrastructure acts as a supporting route on the map, the user has the ability to modify and thus
library for all the other layers and provides consult and purchase tickets offered by travel agencies in the
communication between layers, implements persistence city to different points of interest selected. Each point visited
for business objects. by tourists should be qualified to improve the recommendation
• Application: It contains the business logic where each system in order to improve the recommendations of the user
class represents use cases. In this layer, the providers profile. Similarly, the service offered by the travel agency is
are referenced and implemented with a framework valued.
injection control.
IV. EVALUATION
• Services: This architecture is a powerful constraint on
RESTful Web Services with a decouples server logic A. Evaluation of the tourism sector in the city of Cartagena
from the client. de Indias
To make the diagnosis of innovation processes in ICT for
tourism in the city of Cartagena, we identified 49 travel
agencies with national registry of tourism. However, only 20
companies were selected according to the following criteria:
size agency, recognition, geographical location and
accessibility to information. The travels agencies interviewed
provide tour packages, transport services, hotels, flights,
cruises, corporate travel and travel insurance.

The aim of the survey was to know the level of innovation


of the travel agencies to ensure that the technological
prototype is according with the company’s demand. The
survey reveals that 28.6% of companies have no IT solution.
71.4% of tourism enterprises have technological applications,
of which 53.33% use applications websites where customers
can search for fares, schedules, travel offers, availability in
hotels and holiday destinations; and 46.7% of companies have
static websites to promote tour packages. The survey shows Fig. 3. User Form to classify tourist
that companies surveyed have no mobile applications or The stereotype algorithm iterates through all stereotypes
systems offering personalized services in tourism. Travel available and adds up the likelihood for each property. The
agencies are agreed that software applications must have new resulting stereotypes are presented to the user in Point of
services that integrate with the following variables Interest images (figure 4). Then the user needs to select one of
"innovation", "usability" and "personalization". these images to refine their user profiles and get better
recommendations.
Technological development and innovation are crucial to
the economic growth of regions; however, in the city of
Cartagena, a low investment is evident in science, technology
and innovation. All these elements, lead to the city of
Cartagena as uncompetitive compared to other tourist
destinations.
Low levels of infrastructure, especially technological,
influence the development of tourism, even when the global
trends go to the Internet promotion and an intensive use of
Information Technology and Communications (ICT).

B. 10 Signals
Cartagena de Indias, is a city located in the north of Fig. 4. Point of Interest images to refine user profiles
Colombia and is considered the most important tourist
destination. Cartagena was the main port of America for the
Spanish Empire during the colonial times, and which earned The stereotype algorithm iterates through all stereotypes
their recognition as a Historical and Cultural Heritage of available and adds up the likelihood for each property. The
mankind by UNESCO in 1984. resulting stereotypes are presented to the user in Point of
Interest images. Then the user needs to select one of these
10 Signals is a prototype that shows the planning of images to refine their user profiles and get better
activities in the city of Cartagena de Indias (Colombia), recommendations (Figure 5).
although it could easily be adapted to other destinations and
even contain different destinations. The system identifies
geographical points of interest and draws a tour that connects
these points in routes based on tourist interests.
The first screen (figure 3) of the application is a form in
which users are asked to provide the socio-demographic data
necessary for determining their stereotype (age and gender).
After filling out the form, the application computes the most
relevant stereotypes based on the information that has
previously been provided by the user. The result of this
classification is a degree of matching with respect to each
stereotype through a number between [0, 1] where 1 denotes
perfect match and 0 denotes mismatch.
Fig. 5. Route planning in a city destination when user are classified in at left
one category
C. Validation of the recommendation method an interesting research area nowadays because usually people
are performing in a social rather than individual manner.
Personal data and information about general interests of
tourist is used to evaluate the validity of the user classification The user model of the recommender system in tourism
and the accuracy of the recommendations. Our approach considered: socio-demographic information and Information
requires tourist content data, such as genre, age, interest in about the user’s general interests in tourism to construct the
tourism, and other properties. In the first phase of this stereotypical classes and explicit preferences to estimates on
research, the evaluation is based only on the stereotypical the user’s their interest. The paper also describe how new
model to evaluate the correctness of the stereotypical users are introduced to the system through a stereotype model,
classification and the accuracy of recommendations according in order to make better recommendation.
to the socio-demographic data and POIs dataset.
The model represents a service in the architecture
We recruited 30 testers in Cartagena de Indias City, 15 males proposed according to the Domain Driven Design. We used
and 15 females, with ages ranging from 18 to 45 years with a DDD because it architecture can help handle complex
good technological background. All selected users were behaviors when building software. The proposed architecture,
foreign people. system implementation, recommendation methods, and
algorithms allow implementing intelligent attraction
In order to evaluate the recommender model, we made a information processing that significantly increases the system
satisfaction survey regarding the recommendation they usability.
received. This survey consists of three choice questions, which
have five possible answers: totally disagree, mostly disagree, The proposed architecture, system implementation,
neutral, mostly agree and totally agree. recommendation methods, and algorithms allow implementing
intelligent attraction information processing that significantly
• Q1: It was easy to use this system increases the system usability.

• Q2: the information provided by the system is clear and Acknowledgment


easy to understand This work was supported by the Research Department of
Fundación Universitaria Tecnológico de Comfenalco.
• Q3: I’m satisfied with the POIs suggested

The 75% of the tester says that the app is easy to use, 80% References
think that the app provided clear and easy information and
70% think that they are satisfied with the route suggested. [1] J. Bobadilla, F. Ortega, A. Hernando, A. Gutiérrez, Recommender
During the experiment, evaluators could experiment that the systems survey,Knowledge-Based Systems 46 (2013) 109–132.
suggested recommendations were in line with the tastes of [2] Garcia, I., Sebastia, L., Onaindia, E.: On the design of individual and
group recommender systems for tourism. Expert Syst Appl. (2013)
tourists. However, the application presents issues like gray 38(6):7683–7692.
sheep and cold start, which we intend to fix in later versions. doi:10.1016/j.eswa.2010.12.143.http://www.sciencedirect.com/science/a
rticle/pii/S095741741001506X (2011)
We examine the sensitivity of the results to the number of [3] Moreno, A., Valls, A., Isern, D., Marin, L., & Borràs, J. (2013). Sigtur/e-
destination: ontology-based personalized recommendation of tourism
stereotypes. The performance improves as one increase the and leisure activities. Engineering Applications of Artificial Intelligence,
number of stereotypes. This implies that for recommendation 26(1), 633-651.
systems with frequent introduction of new POIs one should set [4] Bitonto, P.D., Tria, F.D., Laterza, M., Roselli, T., Rossano, V., Tangorra
the number of stereotypes to a relatively large number to F.: Automated generation of itineraries in recommender systems for
achieve higher predictive performance. This provides strong tourism. Proc. International Workshop on Web Engineering and
Tourism (WEBTOUR), Springer Verlag (2010)
evidence towards our system capability to handle unobserved
[5] Ricci, F.: Travel recommender systems. J. IEEE Intelligent Systems 17,
items too, avoid the cold start problem. The classes A, C, D 55-57 (2002)
and F stereotypes provide the best results. [6] Ricci, F., Fesenmaier, D.R., Mirzadeh, N., Rumetshofer, H.,
Schaumlechner, E., Venturini, A., Wöber, K.W., Zins, A.H.: DieToRecs:
V. CONCLUSIONS a case-based travel advisory system. Destination Recommendation
Systems: Behavioral Foundations and Applications. CABI Publisher
Today's tourists demand more information focused to International, Wallingford (2006)
his/her preferences. This paper presents a prototype of a [7] Venturini, A., Ricci, F.: Applying trip@advice recommendation
technology to www.visiteurope.com. In: 17th European Conference on
recommender system in Cartagena City, Colombia, to Artificial Intelligence, pp. 607-611(2006)
recommend a sightseeing tourist route map based on [8] Vansteenwegen, P., Souffriau, W., Berghe, G. V., & Van Oudheusden,
preferences and interest of each user. The model used a KNN D. (2011). The city trip planner: an expert system for tourists. Expert
algorithm to filter and customize the information, in order to Systems with Applications, 38(6), 6540-6546.
generate a Point of Interest list, as a complete plan where user
can follow it during her/his travel. Group recommendation is
[9] Sebastia, L., Garcia, I., Onaindia, E., & Guzman, C. (2009). e-Tourism: [15] Maruyama, A., Shibata, N., Murata, Y., Yasumoto, K., Ito, M.: P-tour: a
a tourist recommendation and planning application. International Journal personal navigation system for tourism. In: 11th World Congress on
on Artificial Intelligence Tools, 18(05), 717-738. ITS, pp. 18–21(2004)
[10] Dunstall, S., Horn, M., Kilby, P., Krishnamoorthy, M., Owens, B., Sier, [16] Hagen, K., Kramer, R., Hermkes, M., Schumann, B., Mueller, P.:
D., Thiebaux, S.: An automated itinerary planning system for holiday Semantic matching and heuristic search for a dynamic tour guide. Proc.
travel. J. Information Technology & Tourism 6, 1-- 33 (2004) Int. Conf. on Information and Communication Technologies in Tourism,
[11] Petrone, G., Ardissono, L., Goy, A.: INTRIGUE: personalized pp. 149 –159 (2005)
recommendation of tourist attractions for desktop and handset devices. J. [17] 18. City Trip Planner: http://www.citytripplanner.com/, accessed 30
Applied Artificial Intelligence 17, 687—714 (2003) January 2011
[12] Goy, A., Magro, D.: Dynamic Configuration of a Personalized Tourist [18] Souffriau, W., Maervoet, J., Vansteenwegen, P., Vanden Berghe, G.,
Agenda. In: IADIS International Conference WWW/Internet 2004, pp. Van Oudheusden, D.: ‘A mobile tourist decision support system for
619- 626. IADIS, Madrid (2004) small footprint devices’. Proc. Tenth Int. Workshop on Artificial Neural
[13] Corchado, J.M., Pavón, J., Corchado, E.S., Castillo, L.F.: Development Networks (IWANN’2009), pp. 1248– 12 55 (2009)
of CBR-BDI agents: A tourist guide application. In: 7th European [19] Horvath, T. : A Model of User Preference Learning for Content-Based
Conference on Case-based Reasoning, pp. 547-- 559. Springer Verlag, Recommender Systems. In Computing and Informatics, Vol. 28, 2009,
Berlin (2004) No. 4, pp. 453–481.
[14] Abbaspour, R.A., Samadzadegan, F.: Itinerary planning in multi-modal [20] Kompan, M., & Bieliková, M. (2013). Personalized recommendation for
urban transportation network. J. Appl. Sci., vol. 9, pp. 1898– 1906) individual users based on the group recommendation principles. Studies
(2009) in Informatics and Control, 22(3), 331-342.

View publication stats

You might also like