Original Scientific Article
Data Mining of Visitors’ Spatial Movement Patterns
Using Flickr Geotagged Photos: The Case
of Dispersed Plečnik’s Architectural Heritage
in Ljubljana
Gorazd Sedmak
University of Primorska, Slovenia
gorazd.sedmak@fts.upr.si
Dejan Paliska
University of Primorska, Slovenia
dejan.paliska@fts.upr.si
Aleksandra Brezovec
University of Primorska, Slovenia
aleksandra.brezovec@fts.upr.si
The aim of this study is to analyse the patterns and structure of spatial visitor be-
haviour in Ljubljana, focusing on the spatially dispersed attractions of Jože Plečnik’s
architectural heritage recently inscribed in the unesco WorldHeritage List.Mean-
ingful incorporation of architectural heritage into the overall tourist experience of
the city poses several challenges for dmos – how to properly communicate the role
and the value of remarkable architectural units, how to regulate uneven visiting times
and place over-concentration, how to provide visitors the opportunity for a rich and
comprehensive tourist experience, and finally, to form ‘cumulative attractions.’ In the
case of Ljubljana, these challenges are compounded by the spatial dispersion of the
elements of the chosen attraction. The objectives of our study were: to illustrate the
spatial interactions between the World Heritage attractions in Ljubljana and their
interaction with other tourist ‘hot spots,’ and to investigate the movement patterns
of visitors to the Plečnik attractions. To this end, Big Data analysis was performed on
geotagged photos uploaded by visitors to the photo-sharing platform Flickr. Spatial
clustering and movement patterns were used to achieve the objectives. The results
show that Ljubljana’s landmarks designed by Plečnik in the old city centre are inte-
grated into a broader attraction network, while the more remote landmarks appear
to be less visited and isolated. It is reasonable to assume that one-day visitors who
have visited one ormore attractions in the historic centre rarely venture further away
and therefore they do not experience theWorldHeritage Site entirely. Themain con-
tribution of this research is a better understanding of the behavioural patterns of dis-
persed unesco site visitors, their structure, and the role of these attractions within
the destination.
Keywords: visitors’ spatial movements, Plečnik’s architectural heritage, big data
analysis, geotagged photos, spatial behavioural patterns
https://doi.org/10.26493/2335-4194.16.49-62
Academica Turistica, Year 16, No. 1, April 2023 | 49
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
Introduction
Understanding inter-destination and especially intra-
destination tourists’ movement patterns is important
for strategic policy-making decisions, organisation of
public transport, planning of road networks and pub-
lic spaces, safety issues, and management and market-
ing of the tourist destination, which includes prod-
uct development and visitor use policies (Caldeira &
Kastenholz, 2020; Lew & McKercher, 2006, 2006; Li
et al., 2019; Park et al., 2020; Vu et al., 2015). While
inter-destination movements have received consider-
able attentionwithin tourism studies (Flognfeldt, 1999;
Oppermann, 1995; Tideswell & Faulkner, 1999), there
have been a relatively limited number of empirical
studies on intra-destination tourist’s movement pat-
terns. One of the main reasons for this is the complex-
ity of tourist movements within the destination, with a
virtually unlimited number of combinations of places
tourists might visit and stochastic individual move-
ment patterns, which make the study of movements
within a destination more challenging compared to
movements between destinations (Mckercher & Lau,
2008). Another challenge is the difficulty in obtaining
relevant and reliable data (Lau & McKercher, 2006).
In the past, data for movement analysis was usually
collected through resource-limited surveys. However,
with the bloom of social media, evolution of mobile
technology and datamining procedures these patterns
have become easier to monitor (Park et al., 2020). In
the last few years, user generated content (ugc) has
become an important source of information for intra-
destinationmovement patterns analysis. In particular,
if we wish to monitor tourists’ movement patterns in
connection to freely accessible tourist attractions or
points of interest, which can be accessed from many
different directions, this approach has strong poten-
tial.
Tourist intra-destination movement is influenced
by a set of destination’s and a set of tourist’s charac-
teristics. The latter include time and money restric-
tions, motivations, transport mode preferences, in-
terests, knowledge, familiarity with the destination,
emotional attachment to the destination or attraction,
etc. (Lew & McKercher, 2006; Zoltan & McKercher,
2015). On the other hand, destination configuration,
type and locations of attractions and accommodation
facilities location, and transportation accessibility –
including costs, congestion and quality of signage and
other destination features – affect tourists movements
as well (Lau & McKercher, 2006).
Architecture as objectified cultural capital is a vi-
tal element of city tourism. It characterises a partic-
ular sense of the place. Especially, ‘the iconic archi-
tecture (buildings, landmarks,monuments) is particu-
larly alluring as it identifies a place’ (Scerri et al., 2016,
p. 1). As such, it has a major role within destination
marketing. The touristic role of urban architecture is
twofold. It can, on the one hand, be seen as a town-
scape, offering pleasant scenery for various touristic
activities or, on the other hand, can represent an attrac-
tion per se. For architecture in urban spaces, Ebejer
(2021, p. 65) suggests the following definition: archi-
tectural attraction is ‘a site that is of sufficient aesthetic,
narrative and cultural interest to provide for the en-
joyment, amusement, entertainment and education of
visitors.’ Understanding the function and relative im-
portance of specific architectural attractions is vital for
their sensible inclusion into the overall tourism prod-
uct and citymarketing activities. This is especially true
in the case of highlight attractions such as unesco
protected buildings.
The aim of this study was to analyse the spatial
movement patterns of visitors in the newly declared
World Heritage city of Ljubljana, focusing on the spa-
tially dispersed attractions of Jože Plečnik’s architec-
tural heritage.
Theoretical Framework
Tourism implies movement, and tourism attractions
spatial distribution undoubtedly plays a crucial role in
shaping tourist spatial movement patterns. As tourists
usually cannot consume all the destination attrac-
tions in a few days’ visit, they have to decide which
attractions they will visit and which not (Shoval &
Raveh, 2004). There are several tourist personal traits
and destination specifics influencing these decisions,
which to a great extent overlap with general factors
defining intra-destination movement. Many schol-
ars have investigated the role of tourists’ socio-demo-
graphic, psychographic and behavioural characteris-
50 | Academica Turistica, Year 16, No. 1, April 2023
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
tics in connection to visited attractions. They found
that tourists with limited time budgets, those with
low incomes, or those travelling in organised groups,
first-time visitors, and foreign tourists tend to visit
only the main attractions of the destination, while
those who are less constrained in terms of time and fi-
nances, individual tourists, repeat visitors, and domes-
tic tourists are more active and explore more and also
more remote areas (Zoltan &McKercher, 2015; Shoval
& Raveh, 2004; Cooper, 1981). Regarding tourist per-
sonality traits, previous studies show that tourists from
the allocentric pole tend to visit and explore a wider
set of attractions compared to more psychocentric
tourists (Debbage, 1991). Another obvious factor af-
fecting the decision on the attractions visit is the length
of stay in the destination (Kang et al., 2018).
On the other side, tourist movement patterns also
depend on the destination’s characteristics, including
number, spatial distribution and density of attractions.
Lue et al. (1996) introduced the concept of cumulative
attraction, where the compatibility of attractions plays
an important role. If there aremany compatible attrac-
tions within an area, they have a greater chance to be
visited than in the case where there is only a single at-
traction in the area (Lue et al., 1996). In addition, pop-
ularity and ratings of attractions cause co-occurrence
of visits between specific attractions (Hernández et al.,
2021). A destination can thus be perceived as a (more
obvious) geographical space or as a relational space (of
attractions), which can be different from each other
(Van der Zee & Bertocchi, 2018). The authors (Van der
Zee & Bertocchi, 2018) stress that the two spaces tend
to be more interrelated for international tourists and
less for domestic ones.Another theory that can explain
the effects of the spatial distribution of attractions on
movement patterns is gravitational theory (Park et al.,
2020). According to this theory, primary attractions
have greater gravitational pull than secondary attrac-
tions, while clustered attractions can create a greater
gravitational effect than a single attraction. Another
approach, the so-called anchor-point theory, was in-
troduced by Couclelis et al. (1987). The so-called an-
chor points refer to primary nodes or reference points
of distinct regions and define the spatial cognition of
individuals. That means that tourists tend to create
their specific cognitivemaps of the destination accord-
ing the relative importance and hierarchical arrange-
ment of attractions (Couclelis et al., 1987).
Movement patterns within the destination are un-
doubtedly also affected by tourists’ specific motives
and affinity for different types of attractions. Spa-
tial distribution of architectural attractions, for ex-
ample, has a greater impact on movement patterns
for cultural tourists than for recreational ones (Steb-
bins, 1996). Therefore, understanding the structure
of tourists and the movement patterns of different
segments in relation to the architectural attractions
is also important for destination marketing decision-
making.
Urban Spatial Structure, Architectural Attractions
and Tourists’ Spatial Behaviour
In the context of tourism, spatial behaviour refers to
the sequence of attractions visited by tourists within
a geographic space and the sequence of movements
between one attraction and another (Caldeira & Kas-
tenholz, 2017). Studies from the field of urban tourism
have confirmed that spatial behaviour and spatial struc-
ture are interdependent (Ashworth, 1988; Karski, 1990;
Law, 1996). Urban spatial characteristics that have
a major impact on tourist spatial behaviour are ‘the
physical configuration of space, the location of attrac-
tions, and the relative distance between accommoda-
tion and attractions’ (Caldeira & Kastenholz, 2020, p.
25).
Architecture has a specific role in urban spatial
structure as it aestheticises spaces with recognisable
markers that create a particular sense of place and
draw tourists into an area by providing a focal point for
tourist attention and experience (Hayllar et al., 2008).
Iconic architecture provides, in the words of Mar-
cus Vitruvius, the great Roman architect and histo-
rian, ‘firmness, utility and delight’ (Scerri et al., 2016).
While firmness refers to structural durability and util-
ity refers to its spatial functionality, delight refers to ar-
chitectural aesthetics. As stated byMaitland and Smith
(2009), architectural aesthetic value is particularly im-
portant to tourists because it involves intense sensory
presence, resonatesmeanings, and expands awareness.
According to Maitland and Smith (2009), the aes-
Academica Turistica, Year 16, No. 1, April 2023 | 51
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
thetics of the built environment influences tourists’
spatial behaviour in three ways. First, the tourist expe-
rience is affected by the built environment design and
the way in which it is consumed. Second, the tourism
experience affects people’s aesthetic judgments and in-
fluences their demands. Third, the form, appearance
and aesthetic qualities of built environments are to
some degree shaped by the desire of cities to impress
visitors (Maitland & Smith, 2009, p. 171).
Urban cultural tourists tend to behave in a spe-
cific way – in order to visit as many historical sites
as possible, they move predominantly in the central
areas and frequently at a fast pace (Caldeira & Kas-
tenholz, 2020, p. 8). Edwards and Griffin (2013) pro-
pose the use of spatial syntax in the analysis of tourists’
spatial behaviour in cities. Space syntax theory, intro-
duced by Hillier and Hanson (1984) and further elab-
orated by Edwards and Griffin (2013), explains spa-
tial relations that consider howdifferent groups organ-
ise and arrange space in which they find themselves.
In their study, Edwards and Griffin (2013) used gps
tracking to find out how various segments of tourists
moved around the cities. Using this method, the au-
thors diagnosed, for example, the lack of spatial dis-
persion of tourists in Sydney and proposed more ef-
ficient wayfinding systems and tourism information
policies. Paulino et al. (2019) noted that, despite the
tendency of tourists to explore areas close or immedi-
ate to their accommodation, touristmovements can be
more concentrated or dispersed due to the influence
of various factors. These include the spatial relation-
ship between attractions, attraction characteristics, ag-
glomeration of attractions, and spatial characteristics
of the destination. Generally, tourists are more willing
to visit remote places if they are unique ormore attrac-
tive (e.g. iconic sights, landmark cultural institutions,
places of historical significance) (Paulino et al., 2019).
Urban intra-destination spatial behaviour can be
examined through movement patterns and multi-
attraction visitation patterns. The former are deter-
mined by territoriality (attractions visited and dis-
tance from accommodation), linearity (patterns of
movement that depend mainly on spatial configura-
tion), locomotion (means of transportation used), and
wayfinding (orientation in physical space), while the
latter refer to intensity (number of nodal points) and
specificity (particular features of attractions) (Caldeira
& Kastenholz, 2020, p. 11).
Over the past two decades, among the multiple
sources of information used to monitor human mo-
bility, location loggers, cell phone satellite position
records, and geotagged content from social media
have been used to track the spatial behaviour of tour-
ists in urban destinations (Domènech et al., 2020).
User-generated content (ugc) has become a central
subject of examination in tourism studies, as users
now produce, share, or tag large amounts of their own
information, including images and videos. Költringer
and Dickinger’s (2015) research shows that ugc is the
richest and most diverse source of online informa-
tion used to analyse tourists’ destination image and
tourists’ spatial behaviour at destinations.
Visualising the geographical positions of photos
taken by tourists is a commonly usedmethod formea-
suring tourist activity in cities, including World Her-
itage cities (cities with unesco World Heritage as-
sets) (e. g. Domènech et al., 2020). Using Big Data,
researchers have tracked tourists and identified areas
of congestion and underutilisation (case of Akka, Is-
rael; Shoval, 2008); compared spatial behaviour pat-
terns of first-time and repeat visitors (case of Hong
Kong; McKercher et al., 2012); measured public use of
iconic buildings (Bilbao case; Plaza et al., 2015) and
identified spatial shift of attention to exceptional ar-
chitecture (Hamburg case; Alaily-Mattar et al., 2022).
Numerous studies have demonstrated the useful-
ness of network analysis in studying tourism system
related networks. In the context of tourist destinations
analysis, more recently, Kádár and Gede (2021) used
network analysis to determine the spatial and tem-
poral complexity of tourist flows in the cross-border
Danube region, Xu et al. (2022) and Jin et al. (2018)
analysed the characteristics of the tourist flow net-
work in Nanjing City̧ Paulino et al. (2019) analysed
the boundaries of destinations, Mou et al. (2020) used
network analysis to study the spatiotemporal changes
of tourist flows in Shanghai, Lozano and Gutierrez
(2018) studied global tourism flows, and Zheng et al.
(2021) analysed the spatiotemporal behaviour of Chi-
nese tourists in the Nordic countries of Europe. These
52 | Academica Turistica, Year 16, No. 1, April 2023
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
authors used various network and node centrality in-
dicators (e.g. degree centrality, weighted degree cen-
trality, betweenness centrality) to assess the network
as a whole and the importance of individual destina-
tions/attractions.
The Case of Plečnik’s Dispersed Architectural
Heritage in Ljubljana
In 2021, the unesco World Heritage Committee in-
scribed the selected works of architect Jože Plečnik
(1872–1957) in Ljubljana in the unesco List of World
Heritage Sites. The unesco WorldHeritage property
consists of a series of dispersed public spaces (squares,
parks, streets, promenades, bridges) and public insti-
tutions (national library, churches, markets, funerary
complex) created in the period between the twoWorld
Wars and sensitively integrated into the pre-existing
urban, natural and cultural context, thus contributing
to the city’s new identity.
As a result of Jože Plečnik’s intervention between
the two world wars, the urban design in Ljubljana has
the easily recognisable characteristics of a symbolic
capital city (unesco, 2021). This is apparent through
the urban landscape design of the two axes: the land
axis and the water axis. The design of both prome-
nades is based on the continuous use of space, which
determines the structure and use of bridges, parks,
squares, markets and other public spaces, as well as
buildings. These public spaces serve as spiritual places
(the churches of St. Michael and St. Francis of Assisi,
Plečnik’s Žale – The Garden of All Saints) and spaces
for relaxation (archaeological park along the Roman
walls and promenades along the embankments of
the Ljubljanica River, Trnovo Quay), as well as en-
abling market activities (Plečnik’s Market), socialising
(Congress Square, the Three Bridges, the Cobblers’
Bridge), and intellectual and cultural activities (Veg-
ova Street, National and University Library). The se-
lection of Plečnik’s works in Ljubljana comprises 14
components:
i The Green Promenade
1. the Congress Square with Zvezda Park
2. Vegova Street
3. the National and University Library
4. the Square of the French Revolution with the
Križanke open-air theatre
ii The Promenade along the Embankments and
Bridges of the Ljubljanica River
5. Plečnik’s Arcades/Market
6. the Three Bridges
7. the Cobblers’ Bridge
8. the Trnovo bridge
9. Trnovo Quay
10. the Sluice Gate
iii Other Plečnik works
11. the Church of St. Michael
12. the Church of St. Francis of Assisi
13. Plečnik’s Žale/the Garden of All Saints
14. The archaeological park/the Roman wall in
Mirje.
The locations of the designated Plečnik’s heritage
in Ljubljana are shown in Figure 1.
Previous studies show that the designation of ar-
chitectural heritage as a World Heritage Site increases
its value in the eyes of tourists, arouses their inter-
est and influences their spatial behaviour (Khairi et
al., 2022). Plečnik’s architecture was a highlight of the
city even before the inscription on the unesco list.
The city dmo ‘Tourism Ljubljana’ was awarded ‘the
best emerging Europe tourism campaign of the year’
in London in 2018. The campaign was based on the
Plečnik heritage. According to dmo staff, the Plečnik
House is one of the most visited tourist spots (Bandur,
2018).
To better understand the role of Plečnik’s heritage
in the tourism system of Ljubljana, questions arise
about the nature of the tourists’ urban experience. As
suggested by Gravari-Barbas (2020), analyses of her-
itage tourism should ‘move away from the heritage
attractions per se to the tourists and their motivations
as constitutive of the visited heritage’ (Gravari-Barbas,
2020, p. 5).
Study Area, Data andMethods
Data Collection and Database Construction
The data for this study has been retrieved from the
photo-sharing platformFlickr (www.flickr.com). Flickr
Academica Turistica, Year 16, No. 1, April 2023 | 53
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
is one of the first and largest social photo sharing plat-
forms to offer a geotagging service, and unlike Insta-
gram, Panoramio, Facebook and others, is available
almost worldwide. There are numerous tourism stud-
ies that use Flickr data, as it is the only major platform
that offers free access to photos and metadata.
Moreover, previous research has demonstrated the
feasibility and reliability of using such data. Su et al.
(2016) analysed the geographical preferences of for-
eign and domestic visitors to China using photos from
the Flickr platform. The authors found an extremely
strong correlation (r = 0.9) between the number of
photos posted and the number of foreign tourists in of-
ficial statistics. The same strong statistical correlation
(r = 0.9) was also found by Kim et al. (2019) in a study
of tourism in protected areas in developing countries
when they compared the average daily number of pho-
tos posted over a year (‘photo-used-day;’ Wood et al.,
2013) with tourism receipts.
An even stronger correlation (r = 0.98) between
the number of photos on the Flickr platform and the
official number of overnight stays in cities along the
Danube was calculated in a study on tourism flows by
Kádár and Gede (2021). Flickr data mining has been
proved effective in previous studies of tourist move-
ment behaviour (for details see e.g. Jankowski et al.,
2010; Vu et al., 2015; Mou et al., 2020; Park et al., 2020;
Kádár & Gede, 2021; Han et al., 2021).
To crawl the data, we used a Python code for recur-
sive Flickr Application Programme Interfaces (api)
calls. The api – flickr.photo.search returns the pub-
licly available photos’ meta-information including the
photo id, photo title, geocode (longitude and lati-
tude), textual tags, photo timestamp, upload date,
owner name and owner id. A boundary box contain-
ing the administrative area of the city was used to limit
a query response.
In the next step, the acquired owner ids were used
to retrieve the information about the user (api –
flickr.people.getInfo), including the user’s name and
location. Then this informationwasmergedwith pho-
tos metadata using the owner id. In this process, we
also downloaded all available photos. However, the
content of the photos was not relevant in this study
and was therefore not included in this analysis. The
collected dataset consisted of the meta-information
of photos taken between January 2007 and Decem-
ber 2018. We were able to crawl 68,520 photo meta-
information records that were taken by 4,735 individ-
ual users. Occasionally, the Flickr apis returned du-
plicate photos, errors, spatial outliers and incomplete
records. These records were deleted from the dataset
during the data cleansing process. The final dataset
contains 65,210 records.
In accordance with the purpose of the study, we
developed a set of rules to distinguish locals (resi-
dents) from tourists. In previous studies (e.g. Kádár
and Gede, 2013; Önder et al., 2016; Su et al., 2016; Li et
al., 2018; Kádár and Gede, 2021) researchers have used
various heuristic methods to identify tourists; how-
ever, none of these methods are completely reliable
or statistically tested. Although our procedure follows
the methods used in previous studies, we applied a
more rigorous classification criterion. Specifically, all
Flickr users who indicated a foreign home country or
a hometown outside Slovenia and for whom the time
span of a sequence of their uploaded photos (within
one year) was less than one month were classified as
tourists. All other users were classified as non-tourists
and thus removed from the dataset.
In the next step, we followed the suggestions of
Hu et al. (2015) and eliminated active user behaviour
bias (caused by users uploadingmultiple photos of the
same micro-location in a very short period of time)
from the data. First, the photo collections of each user
were sorted chronologically. Then, using a spatial and
temporal filter, we merged multiple consecutive pho-
tos of the same attraction/location into a single record.
After cleaning and filtering the dataset, we ended up
with a dataset of 42,572 photos from 3,556 users classi-
fied as tourists. Their spatial distribution is shown in
Figure 1.
The numbers of uploaded photos and users per
year are shown in Table 1.
AOIs Identification and Network Construction
Since in this study we focus on intra-destination tour-
ist movement, which in our case corresponds to tour-
ists’ trajectories from one spatial location to another
(where locations represent tourist attractions or Areas
54 | Academica Turistica, Year 16, No. 1, April 2023
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
Figure 1 Spatial Distribution of Uploaded Photos in Ljubljana
Table 1 Number of Uploaded Photos and Users per Year
Year            
Number of users            
Number of photos            
Of Interest (aois)), the temporal sequence of daily
photos of different users was used to create daily tra-
jectories. In this process, the density-based spatial
clustering algorithm with noise dbscan (Ester et al.,
1996) was applied to identify the most popular aois.
Details of this widely used data mining algorithm for
identifying aois can be found, for example, in Park
et al. (2020), Hu et al. (2015), Vu et al. (2015), or Paliska
et al. (2022). We then aggregated the users’ individual
daily trajectories into daily cluster level (aois) trajec-
tories (see conceptual scheme in Figure 2). In this way,
we constructed aweighted directed networkwhere the
aois (clusters) represent nodes, the edges between
each pair of i and j nodes represent the movements,
and the weight wij of the edges equals the count of
user trajectories (tourist flow) between i and j nodes.
A total of 2,612 trajectories between 559 nodes and
8,537 tourist movements were extracted.
Academica Turistica, Year 16, No. 1, April 2023 | 55
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
1. Spatio-temporal movements
Temporal sequence of user/day
photos (single photos were not
suitable for analysis).
Extraction of 9.229 user’s
trajectories.
2. Identification of AOI
(Areas of Interest)
Identification of point clusters
using density based spatial
clustering (DBSCAN).
3. Spatial aggregation
of trajectories
Point-to-point matrix with
aggregated moves between AOI.
Building network with flows.
4. Tourists
movement
patterns
analysis
Figure 2 Conceptual Scheme of Tourist Trajectories Building Process
Network Analysis
The constructed weighted directed network of tourist
movements provides an opportunity to quantitatively
analyse structural properties of the tourist attractions
and the relations between them. Following the pre-
viously cited studies (e.g. Kádár and Gede, 2021; Liu
et al., 2017), we selected node degree centrality (Free-
man, 1978), node weighted degree centrality (Barrat
et al., 2004; Newman et al, 2004), and betweenness
centrality (Freeman, 1978; Wasserman & Faust, 1994)
to estimate the role and importance of individual at-
tractions. In the context of tourism flows analysis,
node degree (in-degree and out-degree for directed
networks) measures the importance of attractions in
terms of how well (number of edges between nodes)
they are connected to other attractions.
A comparison of node in-degree and out-degree of
each attraction can be used to determine the attrac-
tion’s role in the tourists’ route: as a beginning, core, or
terminal (Shih, 2006). When analysing weighted net-
works, it is common to extend the node degree indi-
cator to the weighted degree. The weighted degree re-
flects the connection frequency (sum of edge weights
or tourist flows) between the target attraction and ad-
jacent attractions. Opsahl et al. (2010) argue that it is
important to consider both indicators when examin-
ing the centrality of a node because node weighted de-
gree only takes into consideration a node’s total level
of involvement in the network and not the number of
adjacent nodes to which it is connected.
The final indicator, betweenness centrality, mea-
sures the number of shortest paths (or weighted short-
est paths) between pairs of non-adjacent nodes that
pass through a given node and reflects the ability of a
given attraction to control interactions between pairs
of other attractions in the attraction network (Shih,
2010). A high betweenness centrality of a particular at-
traction means that tourists would most likely make a
stop at that attraction while travelling between other
attractions (Shih, 2010).
Results and Findings
By visualising the constructed network, valuable in-
sights into the movement can be gained. As can be
seen in Figure 3, movement patterns are spatially con-
centrated within the city centre. Additional analysis of
the non-clustered movement trajectories shows that
nearly two-thirds of the movements (63) occurred
in the city centre between the 10 main attractions
and that the maximum number of moves (213) were
recorded between the Three Bridges area (id1) and the
Robba Fountain area (id3). In addition, the analysis
revealed that more than half (51) of the movements
were related to only two attractions, 15 were re-
lated to three attractions, and 23 of trajectories con-
nected six or more attractions. These results suggest
56 | Academica Turistica, Year 16, No. 1, April 2023
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
Figure 3 Network of Movement between Attractions in Ljubljana
that the destination attracts different tourist profiles –
roughly, those who exhibited a greater spatial move-
ment and are interested in detailed urban exploration,
and those inclined to visit only a limited number of
attractions. Since our network is composed of flows
of one day-trip, differences in movement patterns can
also be attributed to the length of stay at the desti-
nation. Unfortunately, our study was not designed to
investigate those differences. The network also pro-
vides an overview of the popularity of attractions in
terms of the number of photographs taken (Figure 3).
The Three Bridges, together with Prešeren Square, are
themost photographed attractions in Ljubljana (2714),
followed by Ljubljana Castle (1406), Robba Fountain
(1406), Plečnik Market (1357), Dragon Bridge (1322),
Congress Square (1078), and Cobbler’s Bridge (1055),
just to name the places with more than 1000 photos in
the cluster. These attractions are also the most visited
in terms of tourist flows. The analysis of the structural
characteristics of the network provides additional in-
sight into the role and importance of the attractions in
the network (Table 2). Due to space limitations, only
the structural characteristics of the Plečnik works and
the main other attractions are listed in Table 2.
In general, we can see that the attractions in the
old city centre have higher values for node central-
ity and that these values decrease with distance from
the centre. This indicates that centrally located at-
tractions play a dominant role in the network (are
connected with primary flows) and that the impor-
tance of other attractions in tourists’ movement be-
haviour decreases with their distance from the core
attractions. Furthermore, if we compare the values
of in-centrality and out-centrality, no evident differ-
ences emerge. This implies that the attractions are
balanced in the inbound and outbound connections
(in-out degree) and flows (in-out weighted degree),
which means that there are no typical beginning or
Academica Turistica, Year 16, No. 1, April 2023 | 57
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
Table 2 Node Centrality Measures for Selected Attractions in the Network
Attraction () () () () () ()
Three Bridges*      
Cobbler’s Bridge*      
Robba Fountain      
Ljubljana Castle      
The Dragon Bridge      
Congress Square*      
Plečnik’s Market*      
Old Square      
Vegova Street      
National and University Library*      
River banks – Trnovo Quay*      
Skyscraper      
Križanke Open-Air Theatre*      
Slovenian National Drama Theatre      
Riverbanks – Sluice Gate*      
Plečnik’s Žale (area )      
Roman Wall in Mirje (area )*      
Roman Wall in Mirje (area )*      
Trnovo Bridge*      
Plečnik’s Žale (area )*      
The Church of St. Michael*      .
The Church of St. Francis of Assisi*      .
Notes * Plečnik works. Column headings are as follows: (1) attraction id, (2) in-degree, (3) out-degree, (4) in-weighted
degree, (5) out-weighted degree, (6) betweenness centrality.
terminal attractions of the tourist routes. Looking at
Table 2 and Figure 3, we can see that the core attrac-
tions consist of attractions id1, id2, id3, id5, id6,
id7, and id8. According to the values of node central-
ity (node degree, weighted degree and betweenness ),
the most important attraction in the network is the
Three Bridges with Prešeren Square (id1), which is
also the most important stopover that connects pairs
of other attractions.
Because of their popularity, these attractions are in-
cluded in many thematic itineraries. In addition, four
other Plečnik architectural attractions from the un-
esco whs are among the top ten attractions in Ljubl-
jana (in terms of node centrality), namely: Congress
Square (id7), the Plečnik Market (id8), Cobbler’s
Bridge (id2), and the National and University Library
(id11). This is a clear indication that the Plečnik ar-
chitectural heritage is an integral part of Ljubljana’s
tourism system and plays a very important role in the
network of core attractions. Less visited attractions are
peripheral in the network and in blocks with low cen-
trality values. In general, we can observe that as the
distance from the main attractions in the old city cen-
tre increases, all centrality values gradually decrease.
Although it is well known in tourism literature that
spatial flows within a destination are less sensitive to
distance than flows between destinations (Xiao et al.,
2013; Liu et al., 2012; Jin et al., 2018), a significant dis-
58 | Academica Turistica, Year 16, No. 1, April 2023
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
tance decay effect can be observed in tourist move-
ments in our case. Regarding Plečnik architecture, low
centrality values can be observed for some Plečnik
works on the water axis (id126, id63, id29) and for
Plečnik’s architectural attractions at the periphery of
the network (id121, id152, id429, id430).
Discussion and Conclusion
In the research we focused on the behavioural ap-
proach to the spatio-temporal behaviour of tourists
in the city in relation to the architectural heritage of
Jože Plečnik. For this purpose, a set of his works from
the unesco World heritage list have been considered
for interpretation. Movement patterns were identified
by analysing temporal sequences of daily photos re-
trieved from the photo-sharing platform Flickr.
A clearly visible concentration of movements was
identified in the relatively limited area with a high ag-
glomeration of attractions, including the banks of the
Ljubljanica River and the bridges, Vegova Street with
the National and University Library and Congress
Square with Zvezda Park. The area around the three
bridges has proven to be a primary node which, to-
gether with the picturesque Old Town, cultural events,
gastronomic establishments and lively social life, forms
a cumulative attraction (Lue et al., 1996) with a strong
gravitational pull (Park et al., 2020). The friction of
distance related to the three ‘detached’ Plečnik sights
– St. Michael’s Church, the Church of St. Francis of
Assisi, and Plečnik’s Žale – despite relatively good ac-
cessibility, and popularity as well as promotional ex-
posure of Plečnik’s heritage is obviously a more im-
portant factor for (non)visitation than the uniqueness,
iconic character, aesthetic and cultural value of these
attractions (Paulino et al., 2019). Thus, the tourist at-
tractiveness of Ljubljana seems to lie in its overall
value, where no single element stands out. As Hernán-
dez (in Caldeira & Kastenholz, 2020) would put it,
Ljubljana is an ‘attraction city’ rather than a ‘city of at-
tractions.’ Indirectly, these results suggest that there is
a high degree of compatibility of attractions (including
Plečnik’s heritage) around the primary node.
Our findings confirmed intuitive expectations re-
garding the role and degree of integration of Plečnik’s
architectural heritage into the destination’s overall
tourism offerings. Although there is undoubtedly a
segment of tourists primarily interested in the archi-
tectural sites inscribed on the unesco World Her-
itage List, for an average visitor to Ljubljana these at-
tractions seem to represent an organic part of the city’s
picturesque scenery.
The length of stay at the destination, promoted by
Kang et al. (2018) as a factor in attraction visitation
decisions, was not included in the empirical analy-
sis as this is beyond the scope of our work. Still, pre-
liminary research using official statistics data (https://
pxweb.stat.si) suggests that this factor has some in-
fluence on the spatio-temporal behaviour of tourists.
In the summer months, when the average length of
stay is shorter, daily trajectories tend to be shorter,
suggesting that tourists’ movements are less dispersed
and they visit fewer attractions that are further away.
These relations are definitely worth considering in fur-
ther research.
Of course, there are some limitations to the present
study that must be mentioned at the end. One of the
main limitations of this study is that it examines the
movement patterns of tourists in a newly designated
unesco whs. According to previous studies (Khairi
et al., 2022), tourists’ behaviour, and consequently
theirmovement patterns, are expected to changewhen
they become aware of the unesco brand of the city’s
architectural attractions. With this limitation, the re-
sults of this study primarily serve as a situation anal-
ysis that can help destination management take mea-
sures to ensure timely and sustainable management of
tourist flows in the destination.
The small number of photos by domestic visitors
and deficiency and inconsistency of the personal in-
formation disclosed by Flickr users did not allow us
to make comparisons that would show the differences
in movements between segments of domestic and for-
eign tourists.Moreover, the information source itself is
likely to be biased – it is virtually impossible to verify
how representative the sample of Flickr users is in re-
lation to the population of visitors to Ljubljana. Never-
theless, a brief overview of the tags (the most frequent
are: Ljubljana, Slovenia, architecture, Europe, Plečnik,
city, castle, river) suggests that these users are relatively
‘serious’ tourists who focus more on the city’s features
Academica Turistica, Year 16, No. 1, April 2023 | 59
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
than on people or fun. In terms of the potential impli-
cations of the findings on destinationmanagement,we
can note that Ljubljana is already an established desti-
nation with established tourist flows and ‘roles’ of in-
dividual architectural attractions. The organic embed-
ding of Plečnik architecture in the city, defined in the
unesco charter as the central value of its exceptional
world heritage, is indeed reflected in the tourist ‘con-
sumption’ of Ljubljana, which allows for an authentic
and sustainable communication of its exceptionality
with relatively little intervention.
As Plečnik’s scattered attractions were added to the
unesco list only last year, the destination manage-
ment can better prepare for development, promotion
and mobility measures related to Plečnik’s attractions
based on our analyses.We suggest that the destination
be promoted as a ‘new unesco World Heritage City,’
as Plečnik’s attractions are already a key component of
tourist tours. This emphasiswould increase the visibil-
ity and value of the architectural heritage and the city,
especially in the eyes of cultural tourists.
The development of tourism products andmarket-
ing communicationmust also take into account the at-
tractiveness/photogenicity of Plečnik’s attractions and
the existing patterns of their visits/viewing, which can
be seen from the frequencies and sequences of visits of
intra-destination points. Furthermore, it is also useful
to adapt the tourist product andmarketing communi-
cation to different visitor segments (according to av-
erage length of stay, motives and other criteria) and to
the season. To this end, it would be useful to conduct
additional research to identify differences between vis-
itor groups and movement patterns in different parts
of the year.
Finally, the impact of the unesco brand and its
implementation on the spatial behaviour of tourists in
Ljubljana and on the degree of attractiveness of indi-
vidual Plečnik architectural attractions, especially in
remote locations, definitely needs to be measured in
the future with some time lag.
References
Alaily-Mattar, N., Arvanitakis, D., Vordemann, L., & Thier-
stein, A. (2022, 20 June). From exceptional architecture
to city icons? Analyzing data scraped from Flickr. Pro-
jections. https://projections.pubpub.org/pub/nlq0v1z9
/release/2
Ashworth, G. J. (1988). Marketing the historic city for tour-
ism. In B. Goodall & G. J. Ashworth (Eds.), Marketing
in the tourism industry: The promotion of destination re-
gions (pp. 162–175). Croon Helm.
Bandur, S. (2018, 20 August). Turisti imajo radi zlasti Pleč-
nikovo Ljubljano. Delo. https://www.delo.si/magazin
/zanimivosti/turisti-imajo-radi-zlasti-plecnikovo
-ljubljano/
Barrat, A., Barthélémy, M., Pastor-Satorras, R., & Vespig-
nani, A. (2004). The architecture of complex weighted
networks. Proceedings of the National Academy of Sci-
ences, 101(11), 3747–3752. https://doi.org/10.1073/pnas
.0400087101
Caldeira, A., & Kastenholz, E. (2017). Tourists’ spatial be-
haviour in urban destinations: The effect of prior desti-
nation experience. Journal of Vacation Marketing, 24(3).
https://doi.org/10.1177/1356766717706102
Caldeira, A. M., & Kastenholz, E. (2020). Spatiotemporal
tourist behaviour in urban destinations: A framework of
analysis. Tourism Geographies, 22(1), 22–50.
Cooper, C. P. (1981). Spatial and temporal patterns of tourist
behaviour. Regional Studies, 15(5), 359–371.
Couclelis, H., Golledge, R. G., Gale, N., & Tobler, W. (1987).
Exploring the anchor-point hypothesis of spatial cogni-
tion. Journal of Environmental Psychology, 7(2), 99–122.
Debbage, K. G. (1991). Spatial behavior in a bahamian resort.
Annals of Tourism Research, 18(2), 251–268.
Domènech, A., Mohino, I., & Moya-Gómez, B. (2020). Us-
ing Flickr geotagged photos to estimate visitor trajecto-
ries in world heritage cities. isprs International Journal
of Geo-Information, 9(11). https://doi.org/10.3390
/ijgi9110646
Ebejer, J. (2021). Tourism in European cities: The visitor ex-
perience of architecture, urban spaces and city attractions.
Rowman & Littlefield Publishers.
Edwards, D., & Griffin, T. (2013). Understanding tourists’
spatial behaviour: gps tracking as an aid to sustain-
able destination management. Journal of Sustainable
Tourism, 21(4), 580–595.
Ester,M., Kriegel, H. P., Sander, J., &Xu, X. (1996). A density-
based algorithm for discovering clusters in large spatial
databases with noise. kdd’96: Proceedings of the Second
International Conference on Knowledge Discovery and
Data Mining, 96(34), 226–231. aaai.
Flognfeldt, T. (1999). Traveler geographic origin and market
segmentation: The multi trips destination case. Journal
of Travel & Tourism Marketing, 8(1), 111–124.
60 | Academica Turistica, Year 16, No. 1, April 2023
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
Freeman, L. C. (1978). Centrality in social networks concep-
tual clarification. Social Networks, 1(3), 215–239.
Gravari-Barbas, M. (2020). A research agenda for heritage
tourism. Edward Elgar Publishing.
Han, S., Liu, C., Chen, K., Gui, D., & Du, Q. (2021). A
tourist attraction recommendation model fusing spatial,
temporal, and visual embeddings for Flickr-geotagged
photos. isprs International Journal of Geo-Information,
10(1), 20.
Hayllar, B., Griffin, T., & Edwards, D. (2008). City spaces –
tourist places: Urban tourism precincts. Routledge.
Hernández, J. M., Santana-Jiménez, Y., & González-Martel,
C. (2021). Factors influencing the co-occurrence of visits
to attractions: The case of Madrid, Spain. Tourism Man-
agement, 83, 104236.
Hillier, B., &Hanson, J. (1989).The social logic of space.Cam-
bridge University Press.
Hu, Y., Gao, S., Janowicz, K., Yu, B., Li, W., & Prasad, S.
(2015). Extracting and understanding urban areas of in-
terest using geotagged photos. Computers, Environment
and Urban Systems, 54, 240–254.
Jankowski, P., Andrienko, N., Andrienko, G., & Kisilevich,
S. (2010). Discovering landmark preferences and move-
ment patterns from photo postings. Transactions in gis,
14(6), 833–852.
Jin, C., Cheng, J., & Xu, J. (2018). Using user-generated con-
tent to explore the temporal heterogeneity in tourist mo-
bility. Journal of Travel Research, 57(6), 779–791.
Kádár, B. (2014). Measuring tourist activities in cities using
geotagged photography. Tourism Geographies, 16(1), 88–
104.
Kádár, B., & Gede, M. (2013). Where do tourists go? Visual-
izing and analysing the spatial distribution of geotagged
photography. Cartographica, The International Journal
for Geographic Information and Geovisualization, 48(2),
78–88.
Kádár, B., & Gede, M. (2021). Tourism flows in large-scale
destination systems. Annals of Tourism Research, 87,
103113. https://doi.org/10.1016/j.annals.2020.103113
Kang, S., Lee, G., Kim, J., & Park, D. (2018). Identifying the
spatial structure of the tourist attraction system in South
Korea using gis and network analysis. Journal of Desti-
nation Marketing & Management, 9, 358–370.
Khairi, N. D. M., Ismail, H. N., & Jaafar, S. M. R. S. (2022).
Knowledge of tourist spatial behaviour to improveMela-
ka World Heritage Site management. International Jour-
nal of Tourism Cities, 8(1). https://www.emerald.com
/insight/content/doi/10.1108/IJTC-01-2021-0018/full
/html?skipTracking=true
Kim, Y., Kim, C. K., Lee, D. K., Lee, H. W., & Andrada, R. I.
T. (2019).Quantifying nature-based tourism in protected
areas in developing countries by using social big data.
Tourism Management, 72, 249–256.
Költringer, C., & Dickinger, A. (2015). Analyzing destination
branding and image fromonline sources. Journal of Busi-
ness Research 68(9), 1836–1843.
Lau, G., & McKercher, B. (2006). Understanding tourist
movement patterns in a destination: A gis approach.
Tourism and Hospitality Research, 7(1), 39–49.
Law, C. M. (1996). Tourism in major cities. International
Thomson Business Press.
Lew, A., & McKercher, B. (2006). Modeling tourist move-
ments: A Local destination analysis. Annals of Tourism
Research, 33(2), 403–423.
Li, Y., Yang, L., Shen, H., & Wu, Z. (2019). Modeling intra-
destination travel behavior of tourists through spatio-
temporal analysis. Journal of Destination Marketing &
Management, 11, 260–269.
Liu, B., Huang, S. S., & Fu, H. (2017). An application of net-
work analysis on tourist attractions: The case ofXinjiang,
China. Tourism Management, 58, 132–141.
Liu, F., Zhang, J., Zhang, J., Chen, D., Liu, Z., & Lu, S. (2012).
Roles and functions of tourism destinations in tourism
region of South Anhui: A tourist flow network perspec-
tive. Chinese Geographical Science, 22(6), 755–764.
Lozano, S., & Gutiérrez, E. (2018). A complex network anal-
ysis of global tourism flows. International Journal of
Tourism Research, 20(5), 588–604.
Lue, C.-C., Crompton, J. L., & Stewart,W. P. (1996). Evidence
of cumulative attraction inmultidestination recreational
trip decisions. Journal of Travel Research, 35(1), 41–49.
Maitland, R., & Smith, A. (2009). Tourism and the aesthet-
ics of the built environment. In J. Tribe (Ed.), Philosophi-
cal issues in tourism (Aspects of Tourism No. 37, pp. 171–
190). Channel View Publications.
Mckercher, B., & Lau, G. (2008). Movement patterns of
tourists within a destination.TourismGeographies, 10(3),
355–374.
McKercher, B., Shoval, N., Ng, E., & Birenboim, A. (2012).
First and repeat visitor behaviour: gps tracking and gis
analysis in Hong Kong. Tourism Geographies, 14(1), 147–
161.
Mou, N., Yuan, R., Yang, T., Zhang, H., Tang, J. J., & Makko-
nen, T. (2020). Exploring spatio-temporal changes of
city inbound tourism flow: The case of Shanghai, China.
Tourism Management, 76, 103955.
Newman, M. E. (2004). Analysis of weighted networks.
Physical Review E, 70(5), 056131.
Academica Turistica, Year 16, No. 1, April 2023 | 61
Sedmak et al. Data Mining of Visitors’ Spatial Movement Patterns
Önder, I., Koerbitz, W., & Hubmann-Haidvogel, A. (2016).
Tracing tourists by their digital footprints: The case of
Austria. Journal of Travel Research, 55(5), 566–573.
Oppermann,M. (1995). Amodel of travel itineraries. Journal
of Travel Research, 33(4), 57–61.
Opsahl, T., Agneessens, F., & Skvoretz, J. (2010). Node cen-
trality in weighted networks: Generalizing degree and
shortest paths. Social Networks, 32(3), 245–251.
Paliska, D., Kerma, S., & Drobne, S. (2022). Identifying
visitor mobility patterns in Slovenia using Flickr data.
Geodetski vestnik, 66, 175–188.
Park, S., Xu, Y., Jiang, L., Chen, Z., & Huang, S. (2020). Spa-
tial structures of tourism destinations: A trajectory data
mining approach leveraging mobile big data. Annals of
Tourism Research, 84,
Paulino, I., Prats, L., & Schofield, P. (2019). Tourist hub con-
sumption systems: Convenient flexibility versus admin-
istrative constraint. Journal of Hospitality and Tourism
Management, 41, 69–179.
Plaza, B., González-Casimiro, P., Moral-Zuazo, P., & Wal-
dron, C. (2015). Culture-led city brands as economic en-
gines: Theory and empirics. The Annals of Regional Sci-
ence, 54, 179–196.
Scerri, M., Edwards, D., & Foley, C. (2016, 8–11 February).
The value of architecture to tourism [Conference presen-
tation]. 26th Annual cauthe Conference, Blue Moun-
tains International Hotel Management School.
Shih, H. Y. (2006). Network characteristics of drive tourism
destinations: An application of network analysis in tour-
ism. Tourism Management, 27(5), 1029–1039.
Shoval, N., & Raveh, A. (2004). Categorization of tourist at-
tractions and the modeling of tourist cities: Based on the
co-plot method of multivariate analysis. Tourism Man-
agement, 25(6), 741–750.
Shoval,N. (2008). Local icons and the use of space by tourists
in the old city of Akko, Israel. In M. Jansen-Verbeke, G.
K. Priestley, & A. P. Russo (Eds.), Cultural resources for
tourism: Patterns, processes and policies. Nova Science.
Stebbins, R. A. (1996). Cultural tourism as serious leisure.
Annals of Tourism Research, 23(4), 948-950.
Su, M. M., & Wall, G. (2016). A comparison of tourists’ and
residents’ uses of the Temple of Heaven World Heritage
Site, China. Asia Pacific Journal of Tourism Research,
21(8), 905–930.
Tideswell, C., & Faulkner, B. (1999). Multidestination travel
patterns of international visitors to Queensland. Journal
of Travel Research, 37(4), 364–374.
unesco. (2021). World Heritage Convention. https://whc
.unesco.org/en/convention/
Van der Zee, E., & Bertocchi, D. (2018). Finding patterns in
urban tourist behaviour: A social network analysis ap-
proach based on TripAdvisor reviews. Information Tech-
nology & Tourism, 20(1), 153–180.
Vu, H. Q., Li, G., Law, R., & Ye, B. H. (2015). Exploring the
travel behaviors of inbound tourists to Hong Kong using
geotagged photos. Tourism Management, 46, 222–232.
Wasserman, S., & Faust, K. (1994). Social network analysis:
Methods and applications. Cambridge University Press.
Wood, S. A., Guerry, A. D., Silver, J. M., & Lacayo, M. (2013).
Using socialmedia to quantify nature-based tourismand
recreation. Scientific Reports, 3(1), 1–7.
Xiao-Ting, H., & Bi-Hu, W. (2012). Intra-attraction tourist
spatial-temporal behaviour patterns. Tourism Geogra-
phies, 14(4), 625–645.
Xiao, Y., Wang, F., Liu, Y., &Wang, J. (2013). Reconstructing
gravitational attractions ofmajor cities in China from air
passenger flow data, 2001–2008: A particle swarm opti-
mization approach. The Professional Geographer, 65(2),
265–282.
Xu, T., Chen, R., Chen, W., Zheng, L., & Zhang, Y. (2022).
Comparing the spatiotemporal behavior patterns of lo-
cal, domestic and overseas tourists in Beijing based on
multi-source social media big data. Asia Pacific Journal
of Tourism Research, 27(7), 692–711.
Zheng, Y., Mou, N., Zhang, L., Makkonen, T., & Yang, T.
(2021). Chinese tourists in Nordic countries: An analy-
sis of spatio-temporal behavior using geo-located travel
blog data. Computers, Environment and Urban Systems,
85, 101561.
Zoltan, J., & McKercher, B. (2015). Analysing intra-desti-
nation movements and activity participation of tourists
through destination card consumption.TourismGeogra-
phies, 17(1), 19–35.
62 | Academica Turistica, Year 16, No. 1, April 2023