Travellers and Their Joint Characteristics Within the Seven-Factor Model
1. ENTER 2017 Research Track Slide Number 1
Travellers and Their Joint Characteristics
Within the Seven-Factor Model
Julia Neidhardt and Hannes Werthner
E-Commerce Group
TU Wien, Austria
julia.neidhardt@ec.tuwien.ac.at
http://www.ec.tuwien.ac.at
2. ENTER 2017 Research Track Slide Number 2
Outline
• Introduction
• Tourist Types
• Picture-based Recommender System
• The Seven Factors
• The Data Sample
• User Profiles and Individual Characteristics
• User Profiles and Clusters
• Conclusions and Future Work
3. ENTER 2017 Research Track Slide Number 3
Introduction
• Specific challenges for recommender systems in tourism domain
– Tourism product is very complex; average number of items rated per user
is small
– Travelling is an emotional experience, decisions not only rational
– Tourists may not be able to express all their needs; preferences might be
implicitly given
• Personality-based approaches to recommender systems
– Associations between product preferences of users and their personality in
various domains (Tkalcic & Chen, 2015)
4. ENTER 2017 Research Track Slide Number 4
Introduction (2)
• Seven-Factor Model
– User modelled as mixture of factors: Sun & Chill-Out, Knowledge & Travel,
Independence & History, Culture & Indulgence, Social & Sport, Action & Fun,
and Nature & Recreation
– Factors combine short-term behavior (captured by 17 tourist roles) and long-
term personality (captured by the “Big Five” personality traits) (Neidhardt et
al., 2014)
• Aim of this presentation: characterization and statistical analyses of
different groups of users with respect to the seven factors; groups
obtained by:
– Individual attributes of the users
– Machine learning approach
5. ENTER 2017 Research Track Slide Number 5
Tourist Types
• Since 1970s research investigates associations between touristic
behavioral patterns and psychological needs and expectations (Cohen,
1972; Cohen, 1974; Pearce, 1982; Gretzel et al., 2014)
• Framework by Gibson and Yiannakis (2002)
– 17 tourist roles including Action Seeker, Active Sport Tourist, …
– Statistical evidence that touristic behavior is related to psychological needs
and that it changes over time
• Tourist roles can be related to personality traits of travellers, in particular
to “Big Five” model, i.e., extraversion, agreeableness, conscientiousness,
neuroticism, and openness to experience (Jani, 2014; Delic, 2016)
6. ENTER 2017 Research Track Slide Number 6
Picture-based
Recommender System
Picture selection
Travel profile feedback Results
https://www.pixmeaway.com
/
(Neidhardt et al., 2014)
7. ENTER 2017 Research Track Slide Number 7
Knowledge & Travel
The Seven Factors
Independence & HistorySun & Chill-Out
Culture & Indulgence Action & Fun
Nature & RecreationSocial & Sport
8. ENTER 2017 Research Track Slide Number 8
F3: Knowledge
& Travel
The Seven Factors (2)
F2: Independence & History
F1: Sun & Chill-Out
Factor Analysis
Q1 _______
Q2 _______
Q50 ______
Q1 _______
Q2 _______
Q50 ______
Q1 _______
Q2 _______
Q50 ______
Q1 _______
Q2 _______
Q50 ______
Questionnaire addressing
17 tourist roles & “Big Five”
F
4 F
5 F
6
F
7
(Neidhardt et al., 2014)
10. ENTER 2017 Research Track Slide Number 10
User Profiles and
Individual Characteristics
Social & Sports
11. ENTER 2017 Research Track Slide Number 11
User Profiles and
Individual Characteristics (2)
Nature & Recreation
12. ENTER 2017 Research Track Slide Number 12
User Profiles and
Individual Characteristics (3)
Social & SportsIndependence & HistorySun & Chill-Out
13. ENTER 2017 Research Track Slide Number 13
User Profiles and Clusters
• K-means clustering
C1 C2 C2 C4 C5 C6
Size 197 150 179 159 177 135
14. ENTER 2017 Research Track Slide Number 14
User Profiles and Clusters (2)
C1 C2 C3 C4 C5 C6
Sun & Chill-Out ++ ++ 0 ---- + -
Knowledge & Travel -- ++ +++ - -- -
Independence & History -- ++++ - ++ - --
Culture & Indulgence - ++ -- -- - ++++
Social & Sport 0 0 ++ - -- +
Action & Fun ++ + + -- - -
Nature & Recreation ++ ++ 0 ++ --- --
15. ENTER 2017 Research Track Slide Number 15
User Profiles and Clusters (3)
C1 C2 C3 C4 C5 C6
Sun & Chill-Out ++ ++ 0 ---- + -
Knowledge & Travel -- ++ +++ - -- -
Independence & History -- ++++ - ++ - --
Culture & Indulgence - ++ -- -- - ++++
Social & Sport 0 0 ++ - -- +
Action & Fun ++ + + -- - -
Nature & Recreation ++ ++ 0 ++ --- --
Age Group 30 - 39 30 - 39 20 - 29
40 - 49
50 - 59
60 - 0 - 19
16. ENTER 2017 Research Track Slide Number 16
User Profiles and Clusters (4)
C1 C2 C3 C4 C5 C6
Sun & Chill-Out ++ ++ 0 ---- + -
Knowledge & Travel -- ++ +++ - -- -
Independence & History -- ++++ - ++ - --
Culture & Indulgence - ++ -- -- - ++++
Social & Sport 0 0 ++ - -- +
Action & Fun ++ + + -- - -
Nature & Recreation ++ ++ 0 ++ --- --
Age Group 30 - 39 30 - 39 20 - 29
40 - 49
50 - 59
60 - 0 - 19
Gender F M M F
17. ENTER 2017 Research Track Slide Number 17
Conclusions
• Seven-Factor Model as a meaningful and efficient way to represent user
profiles within a computational model
– Factors combine tourist roles and “Big Five”
– Seven factors are capable of distinguishing among different groups and help
to discover patterns
• Group level perspective
– Analyses show that new insights can be gained by switching to group level
(both based on demographics as well as data-driven group detection)
– Can be used for customer segmentation
– Users can be addressed on individual as well as group level (JITT Manifesto)
• Limitation: not enough data in some of the subgroups
18. ENTER 2017 Research Track Slide Number 18
Future Work
• Network level: taking also relationships between the users into
consideration to identify groups and to understand group
characteristics
• Studying cultural differences of travellers with respect to the seven
factors
• User groups and picture selection: how do individual characteristics
and clusters impact picture selection
19. ENTER 2017 Research Track Slide Number 19
The Seven-Factors
https://www.pixmeaway.com
/
20. ENTER 2017 Research Track Slide Number 20
The Seven-Factors (2)
https://www.pixmeaway.com
/
Editor's Notes
Users are not isolated entities but are embedded in a social environment, social context plays a role
Understand this context and use it in our models
Based on seven-factor model, these factors form basis of a seven dimensional vector-space
Factors are represented by 63 pictures
Recommendation process
User selects pictures that he/she finds appealing
Based on the selection, user gets mapped into the seven dimensional space (i.e., he/she obtains a score for each factor)
Also POIs are represented with respect to the seven factors
Recommendations can be delivered based on simple distance function
Sun & Chill-Out, Knowledge & Travel, Independence & History, Culture & Indulgence, Social & Sport, Action & Fun, and Nature & Recreation
Sun & Chill-Out, Knowledge & Travel, Independence & History, Culture & Indulgence, Social & Sport, Action & Fun, and Nature & Recreation
Sun & Chill-Out, Knowledge & Travel, Independence & History, Culture & Indulgence, Social & Sport, Action & Fun, and Nature & Recreation
Gruppeneigenschaften über Profile
Clustering und Segmentierung
Individuell ansprechen und als Gruppe ansprechen
Gruppeneigenschaften über Profile
Clustering und Segmentierung
Individuell ansprechen und als Gruppe ansprechen