ArticlePDF Available

Social Media data: Challenges, opportunities and limitations in urban studies

November 2018
Computers Environment and Urban Systems 74(5)

DOI:10.1016/j.compenvurbsys.2018.11.001

License
CC BY-NC-ND 4.0

Authors:

University of Alicante

Analysing the city through data retrieved from Location Based Social Networks (LBSNs) has received considerable attention as a promising method for applied research. However, the use of these data is not without its challenges and has given rise to a stream of polemical arguments over the validity of this source of information. This paper addresses the challenges and opportunities as well as some of the limitations and biases associated with the collection and use of LBSN data from Foursquare, Twitter, Google Places, Instagram and Airbnb in the context of urban phenomena research. The most recent research that uses LBSN data to understand city dynamics is presented. A method is proposed for LBSN data retrieval, selection, classification and analysis. In addition, key thematic research lines are identified given the data variables offered by these LBSNs. A com- prehensive and descriptive framework for the study of urban phenomena through LBSN data is the main con- tribution of this study.

SMUA's user interface. The left image shows the set up and definition of a search area and the right image shows the search results display page.

…

Google Places. Sub-quadrants derived from the search polygon in compliance with the Google Places API requirements on circle size and maximum number of places retrieved.

…

LBSNs' general data variables.

…

Comparison between Twitter datasets obtained via Streaming and Rest API methods.

…

Instagram API search method.

…

Figures - uploaded by Leticia Serrano-Estrada

Content may be subject to copyright.

Content uploaded by Leticia Serrano-Estrada

Content may be subject to copyright.

Contents lists available at ScienceDirect

Computers, Environment and Urban Systems

journal homepage: www.elsevier.com/locate/ceus

Social Media data: Challenges, opportunities and limitations in urban

studies

Pablo Martí

⁎

, Leticia Serrano-Estrada, Almudena Nolasco-Cirugeda

University of Alicante, Building Sciences and Urbanism Department, Carretera San Vicente del Raspeig s/n. 03690, San Vicente del Raspeig. Alicante, Spain

ARTICLE INFO

Keywords:

Urban planning

Data analysis

Location based social networks

Urban analysis

ABSTRACT

Analysing the city through data retrieved from Location Based Social Networks (LBSNs) has received con-

siderable attention as a promising method for applied research. However, the use of these data is not without its

challenges and has given rise to a stream of polemical arguments over the validity of this source of information.

This paper addresses the challenges and opportunities as well as some of the limitations and biases associated

with the collection and use of LBSN data from Foursquare, Twitter, Google Places, Instagram and Airbnb in the

context of urban phenomena research. The most recent research that uses LBSN data to understand city dy-

namics is presented. A method is proposed for LBSN data retrieval, selection, classiﬁcation and analysis. In

addition, key thematic research lines are identiﬁed given the data variables oﬀered by these LBSNs. A com-

prehensive and descriptive framework for the study of urban phenomena through LBSN data is the main con-

tribution of this study.

1. Introduction

In an era of ever increasing user-generated content —via new data

sources using ﬁxed or mobile sensors such as GPS, credit cards,

smartphones, etc.— patterns of human activity are revealed. Thus,

obtaining meaningful information from these sources represents both a

challenge and an opportunity. Speciﬁcally, digital data sources provide

researchers with a new approach to the study of urban phenomena.

Social media is accepted by scholars as a valuable resource to ad-

vance research on speciﬁc urban aspects (Anselin & Williams, 2015;

Arribas-Bel, Kourtit, Nijkamp, & Steenbruggen, 2015;Roick & Heuser,

2013;Shelton, Poorthuis, & Zook, 2015). Practitioners argue that social

media oﬀers diﬀerent visions on diverse aspects of social, economic and

political urban life reﬂected by the user's interests and activities (Bawa-

Cavia, 2011;Cerrone, 2015;Graham, Hale, & Gaﬀney, 2014;Huang &

Wong, 2015). In fact, the representation and interpretation of data re-

trieved from Location Based Social Networks —LBSNs hereafter—

provide a means by which to assess diﬀerent urban dynamics, such as,

mobility (Cheng, Caverlee, Lee, & Sui, 2011;Luo, Cao, Mulligan, & Li,

2016;Noulas, Scellato, Lambiotte, Pontil, & Mascolo, 2012;Quercia,

Aiello, Schifanella, & Davies, 2015a,2015b); land uses and urban ac-

tivity (García-Palomares, Salas-Olmedo, Moya-Gómez, Condeço-

Melhorado, & Gutiérrez, 2017;Hamstead et al., 2018;Quercia & Saez,

2014;Van Canneyt, Schockaert, Van Laere, & Dhoedt, 2012a,2012b);

human behaviour (Hochman & Manovich, 2013;Lee, Wakamiya, &

Sumiya, 2013;Peña-López, Congosto, & Aragón, 2014;Quercia, Aiello,

Schifanella, & Davies, 2015a,2015b); event detection (Béjar et al.,

2016;Chen & Roy, 2009); and, the issue of adding value to urban

planning, decision making processes and city design (Dunkel, 2015;

Tasse & Hong, 2014).

This research presents an innovative, comprehensive and de-

scriptive method which has been developed for retrieving, processing

and interpreting LBSN geolocated data for the study of cities.

Furthermore, inferences that have been drawn from previous research

are provided to illustrate how this method can be applied to diﬀerent

urban contexts. The novelty of this research lies in the proposed

strategy for addressing the opportunities, limitations and diﬃculties

associated with the process of retrieving, validating, classifying and

ﬁltering LBSN datasets for the study of speciﬁc urban phenomena. Five

LBSNs —Twitter, Foursquare, Google Places, Instagram and Airbnb—

are considered for their unique characteristics and varied metadata to

exemplify these processes.

The paper is structured as follows: i) a recap of the existing litera-

ture on the main opportunities and limitations found in the use of LBSN

data for urban analysis; ii) an explanation of the proposed compre-

hensive data retrieval, selection, ﬁltering and usage method that

overcomes many of the most important drawbacks; and, iii) a discus-

sion of the opportunities and, also, the limitations and diﬃculties of

https://doi.org/10.1016/j.compenvurbsys.2018.11.001

Received 29 May 2018; Received in revised form 23 September 2018; Accepted 2 November 2018

⁎

Corresponding author.

E-mail addresses: pablo.marti@ua.es (P. Martí), leticia.serrano@ua.es (L. Serrano-Estrada), almudena.nolasco@ua.es (A. Nolasco-Cirugeda).

Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

(http://creativecommons.org/licenses/BY-NC-ND/4.0/).

Please cite this article as: Martí, P., Computers, Environment and Urban Systems, https://doi.org/10.1016/j.compenvurbsys.2018.11.001

using LBSN data for the analysis of cities.

1.1. LBSN's in urban analysis: the opportunity

Some inherent features of LBSN data render them a valuable re-

source for developing urban studies.

Firstly, LSBN data are generated by millions of people from diﬀerent

countries throughout the world (Hu et al., 2015) and, as the number of

social network users grows, so does the amount, quality and usability of

data. In 2018, the number of active social media users worldwide

reached 3.196 billion and the number of active mobile social media

users was 2.958 billion (Kemp, 2018).

Secondly, automatic retrieval of social media user-generated con-

tent represents a technological advance for urban analysis. This is

mainly due to the ease with which data collection can be done, re-

moving many of the constraints associated with traditional methods

such as, collection time, accurate geolocation marks, etc. Traditionally,

large surveys and long periods of observation were required to collect

an adequate amount of data for research. Relevant work developed thus

far demonstrates the usefulness of these data for urban analysis. For

instance, some of the most signiﬁcant research that originally used

traditional methods in urban studies: The Image of the city of Boston

(Lynch, 1960) and the Death and Life of Great American Cities (Jacobs,

1961), have been revisited using LBSN data (Al-Ghamdi & Al-Harigi,

2015;Lee et al., 2013;Liu, Zhou, Zhao, & Ryan, 2016;Quercia, Aiello,

Schifanella, & Davies, 2015a,2015b). These studies concur that al-

though closer scrutiny of the data is necessary for more eﬀective data

ﬁltering and mining, crowdsourcing technologies —including social

networks— provide great opportunities for researchers and designers

involved in the analysis of urban environments (Granell & Ostermann,

2016).

Thirdly, the information contained in LBSN data enables the ex-

ploration of intangible aspects of urban life that are linked to places

(McLain et al., 2013). Some social exchanges and events happening in

the city remain concealed (Soja, 1989) in morphological or physical

studies (Cerrone, 2015). However, they leave a virtual trail linked to a

speciﬁc location, which provides a more thorough analysis of users

experiences and perceptions of the city (Saker & Evans, 2016;Silva, Vaz

de Melo, Almeida, Salles, & Loureiro, 2014).

Fourthly, LBSN data can be recognised as volunteered geographic

information —VGI— (Campagna, 2016;Jiang, Alves, Rodrigues,

Ferreira, & Pereira, 2015;Kitchin, 2013) since the expressed percep-

tions, interests, needs and behaviours are published online voluntarily

by the users and refer to unique and speciﬁc places in cities. Data are

generally collected “unobtrusively” (Quercia, Aiello, Schifanella, &

Davies, 2015a,2015b) and users are generally not constrained when

generating information. This is an advantage because according to the

Hawthorne eﬀect, subjects may alter their behaviour in a study on

realization that they are being observed (McCarney, Warner, Iliﬀe, van

Haselen, R Griﬃn, & Fisher, 2007).

Lastly, the diversity of LBSNs, and the content retrieved from them,

oﬀer a multi-perspective approach to the study of cities. There is con-

siderable research using data from Facebook, Twitter and Instagram,

some of the most globally-renowned LBSNs, that covers diﬀerent topics

in relation to diverse ﬁelds of knowledge. However, other LBSNs, such

as Foursquare and Google Places have demonstrated their relevance as

supplementary georeferenced data sources (Jiang et al., 2015;Milne,

Thomas, & Paris, 2012;Serrano-Estrada, Marti, & Nolasco-Cirugeda,

2016;Van Canneyt, Schockaert, et al., 2012a). Moreover, diﬀerent

LBSNs, with the same functionality as the renowned global ones, are

more commonly used in speciﬁc geographical areas. For instance,

Weibo —China— or Mastodon —India— are an alternative to Twitter.

In both South Korea and Japan, Never, is an alternative to Google. Thus,

methods developed for collecting and analysing LBSN data for research

purposes can be transferrable to other LBSNs.

As evidenced by the previously cited works, social media user-

generated data are a valuable by-product for the study of the city

(Arribas-Bel, 2014), and when the information is georeferenced, that

provides added value for urban research since speciﬁc phenomena can

be analysed in a determined urban area. That is the reason why this

study adopts exclusively geolocated data from Social Networks.

1.2. Challenges and limitations of using LBSN data for urban research

Some of the most commonly cited limitations associated with the

use of LBSNs refer to the lack of consistency in the provision of an

acceptable amount of valid geocoded data for each sample (Boyd &

Crawford, 2012;Cerrone, 2015;Leetaru, Wang, Cao, Padmanabhan, &

Shook, 2013;Sloan & Quan-Haase, 2017). For instance, a study con-

ducted in the metropolitan area of Pittsburgh indicated a greatly re-

duced amount of LBSN data generated from urban areas with lower

median income compared to the rest of the city, probably due to lower

smartphone ownership (Tasse & Hong, 2014). Therefore, the amount of

data are largely conditioned by ownership of a smartphone and access

to an internet connection (Arribas-Bel, 2014). Also, there is a diﬀerence

in terms of the quantity of information retrieved from LBSNs between

rural and urban areas (Hecht & Stephens, 2014); and, in the speciﬁc

case of Twitter, the fact that only a small portion of its users activate the

geocoded function when publishing tweets is also an important con-

sideration (Sloan, 2017). Furthermore, the reasons for using the geo-

coding function in Twitter messages are certainly biased by factors such

as social-economic status, political context or education (Graham et al.,

2014). Certain social networks are more popular in some places than in

others, impacting the quantity of information available from a speciﬁc

social network (Sloan & Quan-Haase, 2017). That is why research is

usually applied to case studies involving large metropolitan cities with

a high population density given that there is a considerably greater

amount of LBSN data available for study.

Even if the dataset is acceptable in terms of quantity, lack of

transferability and representativeness in the information provided has

been acknowledged as a problem in the two following circumstances.

Firstly, LBSN data retrieved about speciﬁc locations reveal im-

portant details about the everyday urban life in those places (Lee et al.,

2013;Sui & Goodchild, 2011). Thus, research on single case studies is

limited to a speciﬁc place and it is diﬃcult to know with certainty if the

conclusions obtained from the selected sample are transferrable to

other locations (Goodchild, 2013).

Secondly, there are contrasting opinions about whether LBSNs re-

present the entire population. Some studies argue that LBSN data pro-

vide a representative sample of citizen preferences, opinions and ac-

tivities (Agryzkov et al., 2015;Barbera & Rivero, 2015;Martí, Serrano-

Estrada, & Nolasco-Cirugeda, 2017;Morstatter, Pfeﬀer, Liu, & Carley,

2013;Tufekci, 2014), given the increasing diversity of user proﬁles

(Pew Research Center, 2017). Others claim that LBSN users are not

necessarily a representative sample (Quercia, Aiello, Schifanella, &

Davies, 2015a,2015b) based on the assumption that social media users

comprise only part of the population whose use of a particular social

network tends to be aligned to a speciﬁc interest. However, since no

personal details are retrieved when collecting user data, the sample

cannot be rigorously characterised in terms of user proﬁles as is pos-

sible in a controlled environment —interviews, focus groups, etc.—

(Chorley, Whitaker, & Allen, 2015). Evidently, some users of social

networks are not private individuals but represent organisations, in-

stitutions, businesses, public ﬁgures, and inﬂuencers whose tailored

comments reach and, potentially inﬂuence huge audiences. This case

implies that the data generated is driven by public relations and

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

external communications executives whose role is to comply with the

organization's communications strategy, for example (Cerrone, 2015;

Marwick & Boyd, 2011).

The aforementioned concerns are acknowledged in studies that use

LBSNs for addressing city dynamics. However, the challenges and

limitations associated with the process of data retrieval, veriﬁcation,

selection and ﬁltering have received scant coverage in the literature.

The accuracy of these methods is crucial for obtaining valid datasets

and dealing with diﬀerent research problems concerned with the ﬁeld

of urban studies and this paper seeks to bridge this gap.

2. Method for retrieving and using LBSN data in the study of cities

This section presents a comprehensive method for obtaining, ver-

ifying, ﬁltering and classifying data from ﬁve LBSNs: Foursquare,

Google Places, Twitter, Instagram and Airbnb. Additionally, some in-

ferences are included from previous analysis of urban phenomena using

these data.

2.1. Data retrieval process and tools

There are various methods for retrieving LBSNs data: via Application

Programming Interface —API— (Jagadeesan & Venkatesan, 2015;

Leetaru et al., 2013;Tsou et al., 2013;Wang, 2013;Wilken, 2014; S.

Williams, 2012); via crawled from the website (Mahto & Singh, 2016);

and via purchased by oﬃcial resellers; among others (Mayr & Weller,

2017). Speciﬁcally, this study takes the case of a web-based application

that retrieves data from Foursquare, Google Places, Twitter and In-

stagram: SMUA —Social Media Urban Analyser—. As for Airbnb data, it

is obtained through AirDNA, a third party company that “gathers in-

formation publicly available on the Airbnb website” (AirDNA, 2017).

SMUA's functionality and interface —Fig. 1— has been speciﬁcally

designed to collect geolocated social network data —Table 1—. Laun-

ched in 2013, the ﬁrst version retrieved data from the social networks

Foursquare and Panoramio. However, the latter has been removed after

its closure date on November 4, 2016. Currently, SMUA retrieves data

from Foursquare, Google Places, Twitter and Instagram.

Some aspects involved in the experience with SMUA's retrieval

procedure that are worth highlighting are as follows: ﬁrst, the limita-

tions and requirements imposed by each social network regarding the

shape and size of the search area; and, second, the maximum number of

records provided by the API for each data request. These conditions,

met by SMUA data requests, are commonly found among other LBSNs,

especially those whose data are harvested through an API. Furthermore,

although LBSNs frequently change the requirements in terms of the

number of results per request, the retrieval process remains the same

and, thus could be transferrable to collect data from other similar social

networks. Therefore, the overarching principles of the LBSNs data re-

trieval process through API could be narrowed down to:

1. Request type

2. Search polygon shape

3. Search polygon size

4. Number of requests and/or results allowed per request

5. Timeframe up to data retrieval

6. Retrieved data

These principles, listed in Table 1, have been adopted by SMUA for

each LBSN and will be further explained. It can be observed that Twitter

data can be obtained through two complementary methods: Streaming

and Rest.

The overall procedure for requesting and retrieving data from the

APIs —Foursquare, Google Places, Twitter and Instagram— is as fol-

lows: ﬁrstly, a search polygon area is delineated —of regular or irre-

gular shape— in the Open Street Map cartography (Liftn & Parad, 2018)

within SMUA's interface; secondly, SMUA delineates a Superimposed

Regular Shape —SRS—, rectangular or circular, onto the search

polygon area according to the shape and size restrictions imposed by

the social network's API; and thirdly, the data request is processed. The

data retrieval time will vary according to the size of the SRS. All the

information in the API is retrieved; however, as explained in Section

2.2, a selection, validation and ﬁltering of data is performed before

conducting analysis and drawing any conclusion.

2.1.1. Foursquare and Google Places

The requirements to deﬁne the search area are similar for both

Foursquare and Google Places. Foursquare web service requires the

search area to be a rectangular polygon whose sides cannot exceed the

length of 100 km —Fig. 2— and the maximum number of results pro-

vided per request is 50 records —venues—. Google Places requires the

search area to be a circle, whose radius cannot exceed 5 km in length

—Fig. 3— and the maximum results provided per request is 60 records

—places—.

In commercially active areas, for example, the original search

polygon is very likely to contain more than 50 venues or 60 places.

Therefore, a search algorithm has been incorporated into SMUA to

Fig. 1. SMUA's user interface. The left image shows the set up and deﬁnition of a search area and the right image shows the search results display page.

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

guarantee that all data available from the source is retrieved. The al-

gorithm known as the quadtree decomposition method (Samet, 1984),

which is similar to divide-and-conquer methods (Aho, Hopcroft, &

Ullman, 1974, p. 60), recursively divides the SRS into four quadrants

and, if necessary, the partial quadrants are again subdivided into four

sub-quadrants until the following two conditions are satisﬁed: the

shape sides or circle's radius do not exceed the size limitation set by

Foursquare and Google Places and, concurrently, the number of regis-

ters obtained is less than 50 venues for Foursquare or 60 places for

Google Places —see Fig. 2 and Fig. 3 respectively.

The resulting dataset includes the cumulative list of registers in

Foursquare venues and a list of registered establishments on Google

Places up to the retrieval date.

Table 2 provides details of the number of datapoints retrieved from

Foursquare and Google Places using SMUA from four cities of the

Mediterranean Spanish Arc. The number of datapoints is compared to

the measured area of the continuous urban fabric.

2.1.2. Twitter

Geolocated and non-geolocated tweets can be collected. Twitter

spatiotemporal analyses are conditioned by the amount of data avail-

able taking into account that only part of the tweet traﬃc is geocoded

(Sloan & Morgan, 2015). This is because a tweet geocode can only be

generated from GPS-enabled devices (Han, Cook, & Baldwin, 2014),

and, even though users have full control of whether their tweets are

geolocated or not, the geolocation option in the Twitter app is oﬀ by

default.

There are two ways to include the tweet location: enabling the

precise location of the browser or devise from which the twitter is

broadcast, or select a location label suggested by default by the Twitter

app. In some locations, the latter option includes Twitter places labels

of speciﬁc landmarks, businesses or points of interest that are sourced

from Foursquare (Twitter, 2018a). The presented methodology focuses

on collecting and analysing geolocated tweets retrieved in both ways:

with exact coordinates and those whose location is deﬁned through

Twitter places.

As for retrieving data, there are three diﬀerent ways to access

Twitter API (González-Bailón, Wang, Rivero, Borge-Holthoefer, &

Moreno, 2014): Twitter's Streaming API; Twitter's Search API —Rest

API—; and Twitter's Firehose. SMUA accesses free and open Twitter

data using the ﬁrst two.

The Streaming API is based on real-time data collection. Previous

research has demonstrated that the data search method via Streaming

HTTP protocol using a geographic boundary box as a ﬁlter returns a

very representative sample of tweets (Morstatter et al., 2013).

Twitter's Streaming API requires a rectangular area of any size de-

ﬁned by two pairs of latitude and longitude coordinates. SMUA's al-

gorithm superimposes a SRS and “listens” to the tweets shared within

the deﬁned area. This search method provides a sample of user-geolo-

cated tweets that are occurring real-time within the boundary box. The

Fig. 2. Foursquare. Sub-quadrants derived from the search polygon in compliance with the Foursquare API requirements on size and maximum number of venues

retrieved.

Table 1

Summary of the social networks' API requirements incorporated into SMUA for the data request process.

Foursquare Twitter Google Places Instagram

1. Request type Rest Streaming Rest Rest Rest

2. Search polygon

shape

Rectangular Rectangular Circular Circular Circular

3. Search polygon

size

The sides cannot exceed

100 km

No limitation No limitation Radius cannot exceed

5 km

4. Number of

requests and/or

results allowed per

request

50 results 450 requests per each 15-

min window. No limitation

on the number of results.

Max. 1% of all world-wide generated

tweets.

60 results 5000 calls per hour. No

limitation on the number

of results.

5. Timeframe up to

data retrieval

Venues' cumulative and

updated data

Real time tweets Recently shared tweets —approx. Up to

seven days prior to the retrieval date—

Places' updated data Pictures' updated data

6. Retrieved data Spreadsheet with all venues

registered within the search

area

Spreadsheet with a

representative sample of

tweets collected within the

geographic ﬁlter while

Streaming is activated

Spreadsheet with a listing of tweets is

obtained with no guarantee that all

tweets within the geographic ﬁlter will

be retrieved.

Spreadsheet with all

the places registered

within the search area

photographs tagged

within the area are

retrieved in individual

jpg ﬁles.

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

sample includes those tweets that were shared by the user with a pre-

cise location and those that were tagged with a speciﬁc Twitter place

label. The data collection rate of the Streaming API is limited to 1% of

all world-wide generated tweets (Boyd & Crawford, 2012). Therefore, it

is possible to retrieve all the tweets within a speciﬁc area as long as the

total quantity of tweets requested by the ﬁlter —geographic boundary

box— does not exceed this limitation.

The second method of retrieval, the Rest API, works on requests and

requires the delineation of a circular area with neither a size nor a limit

set for the quantity of results. For the case of SMUA, the limit on the

number of requests is 450 requests per each 15 min window (Twitter,

Inc., 2018b). Despite this data collection method often being used by

researchers (Roberts, 2017;Villatoro, Serna, Rodríguez, & Torrent-

Moreno, 2013), Twitter does not guarantee that the Rest API method

will list all the tweets shared within the search area. In fact, the ﬁnal

dataset per Rest search in Twitter will include a list of tweets that have

been shared in the last seven days approximately.

In both methods, retweets generated by the retweet command on

the Twitter app are not considered original content, and therefore, are

not geolocated. However, copy-pasted tweets generated as new tweets

are considered original content and thus, the user can geolocate them

(Sloan & Morgan, 2015).

A visual comparison of tweets collected using both methods in the

case of Central Park area, New York, —Fig. 4— shows that when re-

quiring the maximum amount of results, the Streaming API method is

preferable, but in terms of obtaining a tweet location pattern over a

short period of time —one week, for example—, the Rest method

provides a random but representative sample. Furthermore, the Rest

method is rather useful in cases where the Streaming API search is not

available due to technical reasons, such as when the internet connec-

tion is interrupted. That said, the combination of both methods would

allow a completer and more accurate dataset.

Fig. 3. Google Places. Sub-quadrants derived from the search polygon in compliance with the Google Places API requirements on circle size and maximum number of

places retrieved.

Table 2

Number of datapoints retrieved from Google Places and Foursquare in relation to the measured area of the continuous urban fabric for four Spanish Mediterranean

Arc cities.

Population within the continuous urban

fabric area (INE, 2011)

Area (km

) (Instituto Geográﬁco

Nacional, 2018)

Google Places Datapoints (data

retrieved: 16 Feb 2018)

Foursquare Datapoints (data

retrieved: 16 Feb 2018)

VALENCIA 782,657 46,68 70,214 15,262

ALICANTE 309,651 35,1 30,758 6417

ELCHE 188,951 19,95 14,880 2633

CASTELLON 153,295 11,94 14,179 2492

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

2.1.3. Instagram

The Instagram data retrieval is conducted by manual and automated

means. The manual method implies downloading photos directly from

Instagram webpage using third-party download plugins, and the auto-

mated download is performed through SMUA. According to previous

research (Boy & Uitermark, 2016;López Baeza, Serrano Estrada, &

Nolasco-Cirugeda, 2016), manual collection is known to have ad-

vantages in terms of the closer analysis of data, especially in qualitative

research where granularity of detail is crucial since data can provide

valuable insights that would not be obtained otherwise from large da-

tasets (Laestadius, 2017). This is because each post is searched and

extracted from Instagram's web service, and more sense can be made of

the pictures in the context of the user's proﬁle page than by using au-

tomated data extraction.

SMUA's automated data retrieval process for Instagram's API search

method consists of a circular shape with a required maximum 5 km

radius. The search area is then covered ﬁrst by a rectangular SRS and

then a circular shape and, if the radius exceeds the allowed distance,

the SRS is subdivided into four sub-quadrants until the circular size

complies with the Instagram API requirements. There is no limitation in

terms of the quantity of registers delivered by the Instagram API.

However, there is a limit of calls per hour which used to be 5000 but

has recently changed to 200 —as of April 2018—.

There are two important diﬀerences between Instagram and the

previously explained three social networks Foursquare, Google Places

and Twitter. Firstly, data retrieved from Instagram —pictures and their

metadata— are not georeferenced to the exact location from where they

were posted. Instead, Instagram has delimited areas with a geolocated

centre point to which all data within the area will be associated

—Fig. 5— For example, all the pictures shared on Instagram in a certain

urban area —namely, downtown area— may be geolocated to the city's

cathedral. Secondly, as of June 2016, Instagram has placed important

restrictions on its API access, one of which is limiting the quantity of

data accessed. Any app or program that intends to retrieve data from

Instagram's API requires system approval ﬁrst. Otherwise, only a

“sandbox” version of the data is available which provides only a very

limited amount of data for retrieval.

2.2. Data variables and usage

Two considerations have an important impact on the analysis of

data retrieved from the social network APIs. Firstly, the LBSNs user-

generated information diﬀers signiﬁcantly from one social network to

the other since they have been designed for diﬀerent purposes. For

instance, users can broadcast their presence by checking-in on

Foursquare venues; register and rate businesses in Google Places, share a

tweet in Twitter, upload a photograph to Instagram and/or comment on

users' images, or register a short-term rental property on Airbnb. Thus,

each social network provides unique data variables —metadata—.

Depending on each research topic, one or several variables from dif-

ferent sources can be considered allowing a more comprehensive as-

sessment process —Fig. 6—. Although data from diﬀerent LBSNs are

not comparable, the resulting analysis from each can be complementary

for research purposes.

Fig. 4. Comparison between Twitter datasets obtained via Streaming and Rest API methods.

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

Secondly, and as a consequence of the previous consideration, re-

ﬂection on the research topic is required prior to selecting suitable data

variables for analysis because not all metadata information oﬀered by

the social network API may be useful for the study of a speciﬁc urban

phenomena. Speciﬁcally, SMUA has been programmed to retrieve only

speciﬁc data variables relevant to the urban phenomena being ad-

dressed. These variables can be grouped into 5 categories, as shown in

Table 3: location [LOC], temporal information [TEMP], user generated

data [UGDAT], data categorization [CAT] and data ID [ID].

The collected data variables have diﬀerent formats depending on

each LBSN: geographic coordinates [coor]; text [txt] —tweets, tips,

comments, hashtags, reviews, photo name or description—; rating va-

lues —check-ins, visitors, rating value— [rat]; photographs [pho];

place,venue or accommodation listing ID [id]; data categories [cat], and

temporal information [temp].

These speciﬁc data variables can be grouped and combined to ad-

dress diﬀerent research topics in the ﬁeld of urban studies. Some of

these potential topics are presented in Section 2.2.2.

2.2.1. Data veriﬁcation and validation

Data harmonization prior to visualization or analysis strengthens

the validity of the data, avoiding errors, over-presence or duplicated

information.

Some users can be extremely active on LBSNs and, thus, skew the

interpretation of the pattern, especially if the size of the sample is re-

duced (Mayr & Weller, 2017). For example, a single user can actively

generate a large number of tweets from a ﬁxed location (Lloyd &

Cheshire, 2017) or, in the case of Foursquare, special promotions ex-

clusive to users checking-in a speciﬁc business establishment may skew

the results for identifying user presence and preference.

Moreover, duplicate venues and places were found in Foursquare and

Google Places, but not in Airbnb.

In the case of Foursquare, duplicate venues can be detected easily

when they have the same name; these kinds of duplicates have been

found to account for less than 10% of dataset listings —i.e. 3.14% in

Prague, Czech Republic; 6.5% in Tallinn, Estonia; and 8.78% in

Valencia, Spain. All datasets retrieved by 11 April 2018—. Other du-

plicate venues that need to be carefully sorted are those that might be a

typo or a diﬀerent name to the same venue and thus be listed twice. For

instance, in the case of the dataset of Alicante, Spain, the venue of the

municipal cemetery is listed twice as: “Cementerio De Alicante”, with

15 check-ins and 12 users; and “Cementerio de Alicante”, with 15

check-ins, 12 users, in dataset retrieved on 16 February 2018. In this

case, both venues are considered as only one and the ﬁnal number of

check-ins and users corresponds to the addition of the two previous

venues.

Fig. 5. Instagram API search method.

Fig. 6. Comprehensive assessment process for the interpretation of data from diﬀerent LBSNs.

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

In Google Places some places are registered twice with a diﬀerent

name. For example, the same restaurant could be referred to as a “bar”

or “cafeteria”. Previous experience has shown that a Google Places

dataset could include up to 2% duplicate listings. One exemplary case is

the raw dataset of the cities of Alicante and Valencia in Spain, with

32,995 and 72,621 place listings, respectively —datasets retrieved on

16 Feb 2018—. After the deletion of duplicate listings, the unique da-

tapoints amounted to 32,392 and 72,019, respectively.

With regard to Twitter and Instagram duplicated data, the duplicate

veriﬁcation of these datapoints is rather simple since every single tweet

and post on Instagram has its own unique ID. However, two relevant

considerations should be taken into account while validating data from

Twitter, especially related to the tweets' locative features. Firstly, copy-

pasted tweets that are generated as new tweets have the same content

with a diﬀerent user ID. In the case where the research requires unique

text, one tweet would need to be deleted. Secondly, when a user or

business is highly active on Twitter, it might skew the analysis of the

spatial tweet pattern distribution as there may be a disproportionate

number of tweets generated from a single location by the same user ID.

Once duplicated data have been removed, perusal of the data before

analysis guides decision making on the appropriate ﬁltering for the

research purpose (Chiera & Korolkiewicz, 2017). The consistency and

organization of datasets largely depend on data categorization, hier-

archy, and structure, which are determined by the LBSN and the users'

criteria for registering and classifying information. Two distinguishable

cases emerge on how data are organized into categories: by tags and/or

user-generated keywords, as in the case of Instagram and Twitter; and

by predetermined categories as in the case of Foursquare, Google Places

and Airbnb.

In the case of LBSNs using keywords to classify the information, data

from Twitter —texts— and Instagram —images— are grouped ac-

cording to the hashtags included in the user's post. The ‘#’—hashtag—

and ‘@’—at— symbols before a keyword or a user allow all posts re-

lated to the same topic or user to be grouped together.

For those LBSNs that use distinctive predetermined categories, such

as Foursquare, Google Places and Airbnb, the validation, reﬁnement

and re-assignment of categories to data is necessary depending on the

research topic and the database's consistency.

Foursquare has 10 general categories (Foursquare Inc., 2017). Each

category is divided into a wide range of sub-categories that provide

more information about the venue's description. Foursquare users re-

gistering a venue on the platform can assign a category and a sub-ca-

tegory; however, the logic behind why some sub-categories are assigned

to a category is not always clear (M. J. Williams & Chorley, 2017).

Although there are some strategies to promote consistency across venue

data (M. J. Williams & Chorley, 2017) —such as a “style guide” and

voluntary reviewers called “Superusers”—, a careful revision of cate-

gories and subcategories is needed.

As for Google Places, when users register a place on the platform,

they assign one or more place types —Google Places categories—. There

are over 120 predeﬁned place types (Google Developers, 2018), thus,

user-assigned categorization of places is even less accurate than in the

case of Foursquare. Therefore, Google Places datasets need to be re-

vised, reﬁned and many places require recategorization prior to analysis

for ﬁve reasons:

(i). As previously mentioned, sometimes places are registered twice

with a diﬀerent name which needs to be considered as one place.

(ii). Some Google Places categories are too general, and/or some places

may not have assigned a speciﬁc sub-category, thus it is not clear

what type of place they represent. Speciﬁcally, the categories

“establishment”, “premise” and “point of interest” could include

all kinds of place types, for instance, restaurants, hotels, oﬃces,

lawyer oﬃces, banks, etc. These places may account for over 32%

of the unique datapoints in a dataset. Taking the previous case

examples, 10,242 listings in Alicante and 22,408 listings in

Valencia, out of the 32,392 and 72,019 unique datapoints, re-

spectively, belong to those three non-speciﬁc categories.

Since

there is a large quantity of these datapoints in the datasets, re-

assigning a category and subcategory is important prior to any

analysis.

(iii). Some data listings do not represent an economic activity or a place

but refer to a larger geographic area or region. That is the case of

places categorised as “street_address”; “postal_town”; and, “sub-

locality_level_4” categories. The number of places that fall within

these categories may represent up to 40% of all unique data list-

ings. Alicante city dataset has 12,557 non-economic activity places

while Valencia has 27,708; out of the 32,392 and 72,019 unique

datapoints, respectively.

(iv). While recategorizing a place, existing Google Places categories

may not be applicable to businesses and places within a speciﬁc

location, thus new categories need to be created. For instance, in

Table 3

LBSNs' general data variables.

General variables Data format/

type

FOURSQUARE TWITTER GOOGLE PLACES INSTAGRAM AIRBNB

1. Location [LOC] Longitude Longitude Longitude Longitude Longitude

coor Latitude Latitude Latitude Latitude Latitude

txt Address, city, country City, country Street, number,

neighbourhood, district,

city, country

Geolocated pin Neighbourhood, city, country

2.Temporal information

[TEMP]

temp Cumulative data on

venues

Time the tweet

was posted

Updated data on

registered places

– Listing creation date

3.User generated data

[UGDAT]

txt Venue name Tweets text Place name Photo description Listing title/ description

num Check-ins – – –

num Users – – – Number of Bedrooms

rat Rating – Rating – Rating

txt Tips, reviews - - Average rate

pho Photographs Photographs – Photographs Photographs

4. Data categorization

[CAT]

cat Hierarchy of categories

and sub-categories

Tweet language

Hashtags

Categories, sub-

categories, sub-sub-

categories

Hashtags Listing type, Property type

—host selects from drop down menu—

5. Data ID [ID] id Venue ID and URL User ID, Tweet ID Place ID Image ID and

URL

Property ID

Since February 16, 2017 some non-speciﬁc general categories such as “es-

tablishment” and “point of interest” have been deprecated (Google Developers,

2018), although places registered prior to that date remain in their originally

assigned categories.

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

the case of the Alicante and Valencia datasets, new categories had

to be created such as “lottery”, with 188 and 478 establishments.

(v). There are cases where the listing location descriptors and ad-

dresses are not homogeneous. For example, in the case of Alicante

and Valencia datasets, the address ﬁeld of 4123 and 719 listings,

respectively, have “Avenida”; 112 and 184 listings respectively

have “Av.”; and, 1172 and 7025 listings respectively have

“Avinguda”. In these cases, harmonization of the terms should be

considered prior to analysis.

Airbnb's temporary accommodation listings are classiﬁed into two

main groups: property type and listing type. These categories have sub-

categories that provide further details about the accommodation char-

acteristics. For example, a property type could be an apartment, bed &

breakfast, boutique hotel, Bungalow, Camper, Dorm, Loft, etc.; and a

listing type refers to whether the accommodation listing includes the

entire apartment, a private room or a shared room. Even though Airbnb

has predeﬁned sub-categories, anybody listing a property can create a

new property type in, for example, a diﬀerent language. For instance,

datasets of many Spanish cities have “casa particular” as a property type.

Thus, as in the case of Google Places, where the listing categories are

reconsidered, the Airbnb's property type categories need to be revised

and possibly grouped into fewer categories —”apartment” and “ser-

viced apartment” listings could fall within the same property type ca-

tegory, for example—.

2.2.2. Data selection, reclassiﬁcation and interpretation

Thorough data variables selection and data reclassiﬁcation, fol-

lowed by detailed examination, is important to ensure that the sample

results are valid for the speciﬁc research purpose and thus can be then

interpreted to obtain representative conclusions (Lansley & Longley,

2016). The variable data selection and classiﬁcation is necessary as the

data have not been generated for urban research purposes. Also, as

previously mentioned, an appropriate selection of data variables is re-

quired, which is conditioned by the research topic to be addressed. The

following Table 4 presents ﬁve example research topics, relevant for the

ﬁeld of urban studies, that will be used to explain ﬁltering methods for

the interpretation of the LBSN data selected: [1] people's perception

and preference over venues can be assessed using Foursquare by ranking

the number of visitors and check-ins and by analysing the venue's user-

shared images and opinions; [2] the diversity and quantity of economic

activities in a speciﬁc urban area can be analysed using Google Places'

listing of businesses; [3] spatiotemporal patterns of people presence,

activities and languages can be assessed using Twitter's geolocated

tweets; [4] the perception and character of the urban environment can

be depicted from the analysis of Instagram images and hashtags; and,

[5] location patterns and building typologies of unregistered temporary

accommodation can be identiﬁed by using data from Airbnb.

Considering the diﬀerent groupings and combinations of data

variables, there are several key points about the ﬁltering methods used

for studying the ﬁve research topics —Table 4— that will subsequently

be dealt with.

2.3. Foursquare: People's perception and preferences

Foursquare datasets include information that is valuable for iden-

tifying people's perception (Quercia, 2015, 2016) and preferences in

cities (Agryzkov, Martí, Tortosa, & Vicent, 2016;Tasse & Hong, 2014;

Van Canneyt, Schockaert, Van Laere, & Dhoedt, 2012b). It is possible to

ascertain the cumulative total amount of visitors and check-ins regis-

tered for each Foursquare venue. Filtering venues by number of visitors

and check-ins allows identiﬁcation of the most visited and thus, the

most preferred venues. However, deciding whether to use the number of

check-ins rather than the number of visitors depends on the research

question itself considering that a single visitor can check-in multiple

times in a venue. Many authors consider the check-ins number for

analysing venue preferences or identifying key points of interest

(Ferreira, Silva, & Loureiro, 2016;Jiang et al., 2015); while other

scholars take the cumulative number of visitors to identify how many

people have checked-in a venue at least once (Bentley, Cramer, &

Müller, 2015;Martí et al., 2017;Noulas, Scellato, Mascolo, & Pontil,

2010). For example, to identify which public plaza is the most socially

relevant in Foursquare, the dataset is ﬁltered so that venues are ranked

according to the number of unique visitors registered under the sub-

category “plaza”, within the general category “outdoors & recreation”

(Martí et al., 2017). This process could be applied to a diﬀerent kind of

venue, for instance, to identify the most preferred restaurants or stores.

Furthermore, the pictures and opinions —tips— shared by

Foursquare users on each venue provide an indication of how the space

is perceived and used (Aliandu, 2015; Y. Chen, Yang, Hu, & Zhuang,

2016). The photographed activities of users —for instance, kids playing

in a plaza— and the urban/architectural features in the background

—fountains, sculptural elements, etc.— could be useful perceptual in-

dicators of a venue's safety for children, or of whether the venue is a

youth-oriented space. However, it is often found that in some cities, the

sharing of pictures and tips is scant because the social network is mostly

intended to showcase presence with check-ins or because the social

network's penetration is low. For example, the case of the most visited

venues categorised as plazas in Foursquare: Plaza Catalunya in Barce-

lona and Plaza Luceros in Alicante. Barcelona has a population of 1.6

million people, and Alicante, a population of just over 300,000. As of 24

August 2018, these plazas had, respectively, 8199 photographs and 656

tips; and, 419 photographs and 41 tips.

2.4. Google Places: The diversity and quantity of economic activities

Information on Google Places listings —classiﬁed by category and

sub-category— reveals clusters of economic activities as well as quan-

tity, diversity, and complexity in the spatial distribution of these ac-

tivities and places of interest. The regrouping of categories into much

fewer and more general categories is helpful not only for making easier

reading and interpretation of cartographies, avoiding the colour coding

Table 4

Example research topics that can be addressed by combining LBSNs data variables.

FOURSQUARE TWITTER GOOGLE PLACES INSTAGRAM AIRBNB

Research topic Identiﬁcation of the

most visited/checked-in

venues. [1]

Spatiotemporal patterns of

people presence, activities

and languages. [3]

Quantity and diversity

of economic activities

in an area. [2]

Identify relevant spatial features/

character related to the user

experience and perception. [4]

location and clusters of accommodation

typology —single family house,

multifamily/ apartment building— [5]

Variables selected UGDAT, ID, TEMP LOC, TEMP LOC, CAT, ID LOC, UGDAT (pho) LOC, CAT

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

of 120 or so place types in Google Places, but also for studying spe-

cialization of economic activities. For instance, previous experiences

have proven that the recategorization of places into the Land Based

Classiﬁcation Standards —LBCS—, speciﬁcally into the “functional di-

mension” hierarchical categories (American Planning Association,

2018), enables the identiﬁcation of location patterns and spatial dis-

tribution of economic activities at diﬀerent scales and granularity. This

classiﬁcation provides a ﬁne-grain land use class taxonomy based on

three levels: 7 main categories, 47 sub-categories and 159 sub-sub-ca-

tegories (Deng & Newsam, 2017). As an example, in the case of Alicante

city dataset —retrieved on 16 Feb 2018—, the allocation of Google

Places place types into the ﬁrst level APA categories resulted as follows:

1000- Residence or accommodation functions- 3.6%.

2000- General sales or services- 46.9%.

3000- Manufacturing and wholesale trade- 4.5%.

4000- Transportation, communication, information and utilities-

9%.

5000- Arts, entertainment and recreation- 14.4%.

6000- Education, public admin, health care and other institutions-

17.12%.

7000- Construction-related businesses- 4.6%.

2.5. Twitter: Spatiotemporal patterns of people presence

In general, selecting variables related to either geo-location or the

tweet content to analyse temporal patterns of activities and people

presence can provide two diﬀerent ﬁltering approaches: spatiotemporal

and/or message content.

Firstly, tweet representations in a cartography by using tweet

timestamps and geolocation is an easy and straightforward way to ob-

serve the concentration patterns of tweets shared in a certain area

(Adnan, Longley, & Khan, 2014;Fujita, 2013;Steiger, Westerholt,

Resch, & Zipf, 2015). The time-based aggregation of tweets could be

useful to understand regular activities happening at a certain hour on a

certain day of the week. For example, in the case of the urban axes that

run and extend along both sides of Paseo de la Castellana in Madrid

—6.3 km long— and the Diagonal avenue in Barcelona —10.2 km

long—, respectively, 61,716 and 14,849 Twitter datapoints were re-

trieved between 21 September 2016 and 17 February 2017. These da-

tasets were aggregated into four daily time periods as shown in Table 5,

showing that daily tweeting patterns in Madrid and Barcelona are quite

similar. This type of ﬁltering process is also applicable to recognise

when one-oﬀ events or demonstrations happen. In the latter case, the

number of tweets increases substantially in a certain urban area and

fades away once the event is ﬁnished (Bolognesi & Galli, 2017;Panteras

et al., 2015).

Secondly, categorization of data by the tweet or the user language

has also proved to be a useful way to identify, for example, the geo-

graphical location of the diﬀerent cultures and nationalities in a city. It

is possible to ﬁnd out what kind and how many foreign languages are

spoken in a certain area (Fisher, 2011;Lange & Waal, 2013). For in-

stance, in the case of Madrid and Barcelona, Spanish in both cities is the

most spoken language among tweeters; however, Barcelona presents a

greater amount of English and “undeﬁned” language tweets, most of

which are in Catalan —Table 5—.

Lastly, the recognition of certain activities, opinions, ideas and

trending topics that are predominant in a given place and at a given

time can be detected by using the information related to the tweet

content —text, hashtags— and sentiment analysis (Cheng et al., 2011;

Yang, Sun, Zhang, & Mei, 2012). Word count techniques can be applied

to a Twitter dataset and represented, for example, in a word cloud using

scaled text size where the higher the frequency of words in a dataset the

larger the font size (Sang & Van Den Bosch, 2013) —Table 5—.

2.6. Airbnb: Location patterns and building typologies of unregistered

temporary accommodation

Airbnb geolocated data provide useful information about the spatial

distribution and concentration patterns of temporary accommodation

by property type or listing type in a given area (Moreno Izquierdo, Ramón

Rodríguez, & Such Devesa, 2016;Temes Cordóvez, Simancas Cruz,

Table 5

Example categorization of Twitter data by daily time periods, tweet language and frequency of hashtags.

Axis Paseo de la Castellana, Madrid Axis Diagonal Avenue, Barcelona

Total tweets collected from 21-09-2016 to 17-02-2017 61,716 14,849

Tweets aggregated by daily time periods

Early morning [0:00 to 6:59 h.] 7.02% 7.85%

Morning [7:00 to 12:59 h.] 27.88% 30.54%

Afternoon [13:00 to 18:59 h.] 36.44% 37.03%

Evening-night [19:00 to 23:59 h.] 28.66% 24.58%

Tweet languages

Spanish 68.55% 35.84%

English 17.34% 25.91%

Italian 0.60% 1.61%

Portuguese 3.89% 1.54%

Undeﬁned 4.43% 27.50% Mostly Catalan

Others 5.19% 7.60%

Word cloud with most repeated hasthags.

P. Martí et al.

Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

Peñarrubia Zaragoza, Moya Fuero, & García Amaya, 2016). The deﬁ-

nition of the diﬀerent accommodation categories is vague —for ex-

ample, it is diﬃcult to know the diﬀerence between the property types

service apartment vs. apartment— and the user-generated information

is not homogeneously classiﬁed. Therefore, it becomes necessary to

regroup property type categories. For instance, in a study conducted of 9

Spanish cities —Alicante, Benidorm, Calpe, Castellón, Gandía, Pe-

ñíscola, Teulada, Torrevieja and Valencia—, a new categorization of

listings by property type was proposed to specify the listing's building

typology: i) multifamily housing apartment; ii) single family housing,

iii) private room, and iv) others —Table 6.

2.7. Instagram: The perception and character of the urban environment

Instagram data oﬀer relevant insights about what is interesting in

the urban environment for people. Pictures shared through Instagram

“promote visual rather than textual communication” (Laestadius,

2017), thus the analysis of the character and identity of the urban en-

vironment can be depicted from the ﬁltering and studying of a much

smaller dataset than in other types of data. However, unlike the pre-

viously explained social networks whose information is retrieved in the

form of a spreadsheet, ﬁltering large sets of Instagram images can be

rather challenging and still remains largely inaccessible for researchers

(Laestadius, 2017). There are open tools available that can classify

pictures automatically according to their hue and luminosity (Hochman

& Manovich, 2013;Manovich, 2016). These techniques are useful to

identify, for example, which pictures are taken indoors or outdoors and

to gauge the extent to which users are interested in outdoor and/or

indoor activities.

The manual ﬁltering and geocoding of pictures retrieved via screen-

captured posts (Laestadius, 2017) and/or Instagram webpage down-

loads (López Baeza, Serrano Estrada, Nolasco-Cirugeda, Serrano-

Estrada, & Nolasco-Cirugeda, 2016) is often done by using place

hashtags. These correspond to a geolocated point that represents a place

or a region —i.e. #centralpark; #newyork—, thus ﬁltering by these

hashtags is a straightforward way to obtain a sample with images that

are shared in a speciﬁc urban location. Another type of picture ag-

gregation and ﬁltering is done by categorizing the content of the pic-

ture; for instance: a selﬁe; a person posing nearby a speciﬁc urban

element —tree, monument—; landscape; scenery; and, architecture.

Moreover, ascertaining people's activities in photos could provide an

indication of the perception of the surrounding space.

2.8. Other research topics

A compilation of other potential research topics using the afore-

mentioned social networks and their respective data variables are listed

in Table 7.

Table 6

Airbnb's property type categories grouped into building's typology classiﬁca-

tion.

Multifamily Single family Private room Others

Apartment Bungalow Bed & breakfast Camper

Boutique hotel Cabin Casa particular Boat

Condominum Chalet Dorm Igloo

Entire ﬂoor House Guest suite

Loft Villa Guest house

Timeshare Townhouse Hostel

Others Nature lodge Timeshare

Serviced apartment Earth House In law

Vacation home

Table 7

Potential research topics in the ﬁeld of urban studies that can be approached by using LBSN data variables.

FOURSQUARE TWITTER GOOGLE PLACES INSTAGRAM AIRBNB

Researchtopic Oﬀer of economic activities and venues of

public interest

People presence in the urban public or private

space.

Public opinion/evaluation of a business or

service.

Keywords/ hashtags related to the user

experience/opinion of a place

location and clusters of the

diﬀerent residential rental types

—single room, entire property—

Variables selected LOC, UGDAT (txt, rat, pho), ID LOC UGDAT (rat), ID LOC, UGDAT (txt) LOC, CAT

Research topic Cumulative people presence in a venue up

to the retrieval date.

Text and/or hashtags to depict user location

—district, neighbourhood, city—

Particularities of the case study derived from

the business oﬀer. Economic activities and

services that characterize an urban area.

Identify the social/public activities

developed in a space.

Geographical distribution of

rental homes with their respective

rating values.

Variables selected LOC, UGDAT (rat), ID LOC, UGDAT (txt) LOC, CAT, ID LOC, UGDAT (pho) LOC, UGDAT (rat)

Research topic Tips, reviews or comments —public

opinion—

Depict cultural features, traditions, routines,

habits of residents through the text they share.

Economic activities on the main ﬂoor that

contribute to the livability of urban spaces.

User Proﬁle of frequenters—gender,

approximate age—

Average rating value of rental

homes for selected urban areas.

Variables selected UGDAT (txt) LOC, UGDAT (txt) LOC, CAT, ID UGDAT (pho) LOC, UGDAT (rat)

Research topic Type of activity that takes place in the

space.

Frequency of people tweeting in an urban area. Predominant economic activities and

specialization of a neighbourhood.

Description of a space through hashtags

as keywords

Physical qualities of the best rated

rental types.

Variables selected LOC, CAT, ID, UGDAT (txt) LOC, TEMP, ID LOC, CAT, ID UGDAT (txt) UGDAT (pho, rat)

Research topic Physical characteristics of the venues

relevant to user experiences

Opinion, emotions about relevant events, social

and political matters.

Variables selected UGDAT (pho) LOC, UGDAT (txt)

Research Topic Local habits and social behaviour in the

space.

Opinion, perception about urban spaces.

Variables selected LOC, GDAT(pho) LOC, UGDAT (pho)

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

3. Discussion and conclusions

The ﬁndings of this research back the many previously cited urban

scholars who support the use of LBSN data for the study of cities. This

trend is set to continue given that content generated by an ex-

ponentially growing community of LBSN users cannot be neglected in

urban research of a qualitative nature. These data can potentially

trigger more discussion about current trends in urban reality than tra-

ditional sources, which cannot compete in terms of immediacy, avail-

ability and quantity of data.

This study underscores the importance of addressing the challenges,

limitations as well as the opportunities provided by LBSN data for the

ﬁeld of urban studies. A new framework is presented in this study for

overcoming several challenges associated with the retrieval, validation,

selection, ﬁltering and interpretation of geolocated user-generated data

—from Twitter, Foursquare, Google Places, Instagram and Airbnb—.

The ﬁndings evidence that a close review and manual veriﬁcation are

required to avoid losing the implicit nuances of each dataset and

thereby, of each case study.

Furthermore, two issues may compromise the rigorous procedure

and the reproducibility of this type of research. First, reliance on data

accessibility makes the retrieval process vulnerable to the changes in

access conditions; and, second, the excessive amount of data makes

manual veriﬁcation of large datasets impractical and implies certain

automatization processes —a script, for example—. Accordingly, the

increasing availability of free and open data means a more re-

presentative sampling, but, as argued by Boyd & Crawford (2012)

“bigger data are not always better data”.

LBSN-oriented methods present limitations for the study of cities in

terms of the representability and applicability of the data, according to

some previously cited scholars. This research recognizes the constraints

associated with using LBSN data for the analysis of urban phenomena,

with speciﬁc reference to: [1] the complexity involved in requesting

and retrieving data according to each LBSN; [2] the amount of data

retrieved, whether the sample is too small to be representative or too

large to manage; [3] the validation, selection, ﬁltering and interpreta-

tion of data, as a process that is conditioned by the complexity of the

research topic and the distinctive variables obtained from each social

network.

In relation to the complexity involved in requesting and retrieving

data [1], this research underscores the importance of dealing properly

with the API requirements in terms of shape and size of the search

polygon and the number of results per request. Precisely,

one of the main methodological contributions is the recognition of

key aspects involved in the data retrieval, making the method trans-

ferrable to other LBSNs. For example, even though most common social

media APIs use the Rest method (Brown, Soto-Corominas, Suárez, & de

la Rosa, 2017), an approach to Twitter Streaming API request method is

rather similar not in terms of quantity but in terms of data represent-

ability, as explained in Section 2.1.2 Twitter, Fig. 4.

As for the number of datapoints retrieved [2], the information from

a speciﬁc location can be far richer if the resulting analyses of two or

more sources are considered when approaching a single research case

study. There are some aspects that can be related among social net-

works such as the relation between the number of datapoints and the

measured area of cities —Table 2—, or apparently common types and

formats of data variables —such as venues and places—. However, the

raw data from two diﬀerent sources should not be compared until the

data have been independently analysed—Fig. 6—. These analysed data

can be complementary to address a research topic. For example,

Foursquare and Google Places both provide a listing of points of in-

terest. However, the size of the dataset, the data variables and the

purpose for which users share data is rather diﬀerent.

That is why the veriﬁcation and selection [3] processes are im-

portant as they may show that there are places registered in Google

Places that are not present in Foursquare and vice versa. In this case, a

business or urban area may not be considered relevant by Foursquare

users —not checked-in—, but the establishment may be listed in Google

Places. Similarly, a recently opened venue may not yet be listed in

Google Places, but it may have check-ins on Foursquare. Thus, the

combination of the resulting analyses of ﬁltered and selected data from

diﬀerent LBSNs can supplement the information on a sample to produce

a more complete and accurate research approach.

Comprehensive research on a speciﬁc urban topic may require the

consideration of validated information from diﬀerent LBSN datasets

and, therefore, the selection of diﬀerent variables. Notably, analysing

variables related to images —Instagram, Twitter and Foursquare— re-

mains challenging in terms of the slowness of the procedure. Although

advances in image recognition software can facilitate this task, each

image still needs to be viewed manually to appreciate local nuances

(Boy & Uitermark, 2016) related to social activity, for example.

Finally, the main contribution of this work is a comprehensive fra-

mework for the study of cities that eﬀectively deals with the challenges

and opportunities provided by readily accessible user-generated LBSN

data. The approach presented could beneﬁt urban design and planning

intervention criteria.

Acknowledgements

This work was supported by the Council of Education, Research,

Culture and Sports – Generalitat Valenciana (Spain). Project: Valencian

Community cities analysed through Location-Based Social Networks

and Web Services Data. Ref. no. AICO/2017/018.

References

Adnan, M., Longley, P. A., & Khan, S. M. (2014). Social dynamics of Twitter usage in

London, Paris, and New York City Citation Format. First Monday, 19(5).

Agryzkov, T., Nolasco-Cirugeda, A., Oliver, J. L., Serrano-Estrada, L., Tortosa, L., &

Vicent, J. F. (2015). Using data from Foursquare Web Service to represent the

commercial activity of a city. International Journal of Computer, Control, Quantum and

Information Engineering. World Academy of Science, Engineering and Technology, 9(1),

69–76.

Agryzkov, T., Martí, P., Tortosa, L., & Vicent, J. F. (2016). Measuring urban activities

using Foursquare data and network analysis: A case study of Murcia (Spain).

International Journal of Geographical Information Science, 1–22.

Aho, A. V., Hopcroft, J. E., & Ullman, J. D. (1974). The Design and Analysis of Computer

Algorithms. Reading: Addison-Wesley Publishing Company.

AirDNA (2017). Short-Term Rental Data Methodology - The AI and science behind

AirDNA. Retrieved 28 August 2018, from https://www.airdna.co/methodology.

Al-Ghamdi, S. A., & Al-Harigi, F. (2015). Rethinking image of the City in the Information

Age. Procedia Computer Science, 65, 734–743. https://doi.org/10.1016/j.procs.2015.

09.018.

Aliandu, P. (2015). Sentiment Analysis to Determine Accommodation, Shopping and

Culinary Location on Foursquare in Kupang City. Procedia Computer Science, 72,

300–305. https://doi.org/10.1016/j.procs.2015.12.144.

American Planning Association (2018). LBCS Function Dimension with Descriptions.

Retrieved 18 January 2018, from https://www.planning.org/lbcs/standards/

function.htm.

Anselin, L., & Williams, S. (2015). Digital Neighborhoods.

Arribas-Bel, D. (2014). Accidental, open and everywhere: Emerging data sources for the

understanding of cities. Applied Geography, 49, 45–53. https://doi.org/10.1016/j.

apgeog.2013.09.012.

Arribas-Bel, D., Kourtit, K., Nijkamp, P., & Steenbruggen, J. (2015). Cyber Cities: Social

Media as a Tool for Understanding Cities. Applied Spatial Analysis and Policy, 8(3),

231–247. https://doi.org/10.1007/s12061-015-9154-2.

Barbera, P., & Rivero, G. (2015). Understanding the Political Representativeness of

Twitter users. Social Science Computer Review, 33(6), 712–729. https://doi.org/10.

1177/0894439314558836.

Bawa-Cavia, A. (2011). Sensing the urban: using location-based social network data in

urban analysis. Pervasive PURBA Workshop (pp. 1–7). .

Béjar, J., Álvarez, S., García, D., Gómez, I., Oliva, L., & Tejeda, A. (2016). Discovery of

spatio-temporal patterns from location-based social networks. Journal of Experimental

& Theoretical Artiﬁcial Intelligence, 28(1–2), 313–329. https://doi.org/10.1080/

0952813X.2015.1024492.

Bentley, F., Cramer, H., & Müller, J. (2015). Beyond the bar: The places where location-

based services are used in the city. Personal and Ubiquitous Computing, 19(1),

217–223. https://doi.org/10.1007/s00779-014-0772-5.

Bolognesi, C., & Galli, A. (2017). Mapping Socials a Voluntary Map of a Great Event in

Monza Park. Proceedings.Vol. 1.Proceedings (pp. 917–). . https://doi.org/10.3390/

proceedings1090917.

Boy, J. D., & Uitermark, J. (2016). How to Study the City on Instagram. PLoS One, 11(6),

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

e0158161. https://doi.org/10.1371/journal.pone.0158161.

Boyd, D., & Crawford, K. (2012). Critical questions for big data: Provocations for a cul-

tural, technological, and scholarly phenomenon. Information Communication and

Society, 15(5), 662–679. https://doi.org/10.1080/1369118X.2012.678878.

Brown, D. M., Soto-Corominas, A., Suárez, J. L., & de la Rosa, J. (2017). Overview- the

social media data processing pipeline. In A. Q.-H. Luke Sloan (Ed.). The SAGE

Handbook of Social Media Research Methods (pp. 125–145). London: SAGE

Publications Ltd.

Campagna, M. (2016). Social Media Geographic Information: Why social is special when

it goes spatial? European Handbook of Crowdsourced Geographic Information (pp. 45–

54). .

Cerrone, D. (2015). A Sense of Place. Turku.

Chen, L., & Roy, A. (2009). Event detection from ﬂickr data through wavelet-based spatial

analysis. Proceedings of the 18th ACM Conference on Information and Knowledge

Management (pp. 523–532). . https://doi.org/10.1145/1645953.1646021.

Chen, Y., Yang, Y., Hu, J., & Zhuang, C. (2016). Measurement and analysis of tips in

foursquare. 2016 IEEE International Conference on Pervasive Computing and

Communication Workshops, PerCom Workshops 2016 (pp. 4–7). .

Cheng, Z., Caverlee, J., Lee, K., & Sui, D. Z. (2011). Exploring millions of Footprints in

Location sharing Services. Icwsm, 2010, 81–88.

Chiera, B. A., & Korolkiewicz, M. W. (2017). Visualizing big Data: Everything Old is New

again. In F. P. García Márquez, & B. Lev (Eds.). Big Data ManagementSpringer

International Publishinghttps://doi.org/10.1007/978-3-319-45498-6_1.

Chorley, M. J., Whitaker, R. M., & Allen, S. M. (2015). Personality and location-based

social networks. Computers in Human Behavior, 46, 45–56. https://doi.org/10.1016/j.

chb.2014.12.038.

Deng, X., & Newsam, S. (2017). Quantitative Comparison of Open-Source Data for Fine-

Grain Mapping of Land Use. Proceedings of the 3rd ACM SIGSPATIAL Workshop on

Smart Cities and Urban Analytics - UrbanGIS.Vol. 17.Proceedings of the 3rd ACM

SIGSPATIAL Workshop on Smart Cities and Urban Analytics - UrbanGIS (pp. 1–8). .

https://doi.org/10.1145/3152178.3152182.

Dunkel, A. (2015). Visualizing the perceived environment using crowdsourced photo

geodata. Landscape and Urban Planning, 142, 173–186. https://doi.org/10.1016/j.

landurbplan.2015.02.022.

Ferreira, A. P. G., Silva, T. H., & Loureiro, A. A. F. (2016). Beyond Sights: Large Scale

Study of Tourists' Behavior using Foursquare Data. Proceedings - 15th IEEE

International Conference on Data Mining Workshop, ICDMW 2015 (pp. 1117–1124). .

https://doi.org/10.1109/ICDMW.2015.234.

Fisher, E. (2011). Language communities of Twitter. Retrieved 20 July 2001, from

https://ﬂic.kr/p/ayDr8X.

Foursquare Inc (2017). Foursquare Category Hierarchy. Retrieved 1 January 2018, from

https://developer.foursquare.com/docs/resources/categories.

Fujita, H. (2013). Geo-tagged Twitter collection and visualization system. Cartography and

Geographic Information Science, 40(3), 18. https://doi.org/10.1080/15230406.2013.

800272.

García-Palomares, J. C., Salas-Olmedo, M. H., Moya-Gómez, B., Condeço-Melhorado, A.,

& Gutiérrez, J. (2017). City dynamics through Twitter: Relationships between land use

and spatiotemporal demographics Cities. https://doi.org/10.1016/J.CITIES.2017.09.

007.

González-Bailón, S., Wang, N., Rivero, A., Borge-Holthoefer, J., & Moreno, Y. (2014).

Assessing the bias in samples of large online networks. Social Networks, 38(1), 16–27.

https://doi.org/10.1016/j.socnet.2014.01.004.

Goodchild, M. F. (2013). The quality of big (geo)data. Dialogues in Human Geography,

3(3), 280–284. https://doi.org/10.1177/2043820613513392.

Google Developers (2018). Place Types. Retrieved 30 April 2018, from https://

developers.google.com/places/supported_types.

Graham, M., Hale, S. A., & Gaﬀney, D. (2014). Where in the world are you? Geolocation

and Language Identiﬁcation in Twitter. The Professional Geographer, 1–11.

Granell, C., & Ostermann, F. O. (2016). Beyond data collection: Objectives and methods of

research using VGI and geo-social media for disaster management. Computers,

Environment and Urban Systems, 59, 231–243. https://doi.org/10.1016/j.

compenvurbsys.2016.01.006.

Hamstead, Z. A., Fisher, D., Ilieva, R. T., Wood, S. A., McPhearson, T., & Kremer, P.

(2018). Geolocated social media as a rapid indicator of park visitation and equitable

park access. Computers, Environment and Urban Systems.https://doi.org/10.1016/j.

compenvurbsys.2018.01.007.

Han, B., Cook, P., & Baldwin, T. (2014). Text-based twitter user geolocation prediction.

Journal of Artiﬁcial Intelligence Research, 49, 451–500. https://doi.org/10.1613/jair.

4200.

Hecht, B., & Stephens, M. (2014). A Tale of Cities: Urban Biases in Volunteered Geographic

Information. Icwsm. 197–205. http://doi.org/papers3://publication/uuid/B13C63A5-

B3B8-4619-9558-86BCAFE5E2CA.

Hochman, N., & Manovich, L. (2013). Zooming into an Instagram City: Reading the local

through social media.

Hu, Y., Gao, S., Janowicz, K., Yu, B., Li, W., & Prasad, S. (2015). Extracting and under-

standing urban areas of interest using geotagged photos. Computers, Environment and

Urban Systems, 54, 240–254. https://doi.org/10.1016/j.compenvurbsys.2015.09.

001.

Huang, Q., & Wong, D. W. S. (2015). Modeling and Visualizing regular Human Mobility

patterns with uncertainty : An example using Twitter Data Modeling and Visualizing

regular Human Mobility patterns with uncertainty : An example using Twitter Data.

Annals of the Association of American Geographers, 105(6), 1179–1197 November.

INE (2011). Instituto Nacional de Estadística. Retrieved 11 May 2018, from http://www.

ine.es/censos2011_datos/cen11_datos_resultados.htm.

Instituto Geográﬁco Nacional (2018). Centro Nacional de Información Geográﬁca.

Retrieved 4 April 2018, from http://www.ign.es/web/ign/portal/inicio.

Jacobs, J. (1961). The death and life of great American cities. New York: Vintage Books.

Jagadeesan, J., & Venkatesan, N. (2015). Study of API for web applications. International

Journal of Contemporary Research in Computer Science and Technology, 1(7), 257–261.

Retrieved from http://www.ijcrcst.com/papers/IJCRCST-OCTOBER15-07.pdf.

Jiang, S., Alves, A., Rodrigues, F., Ferreira, J., & Pereira, F. C. (2015). Mining point-of-

interest data from social networks for urban land use classiﬁcation and disaggrega-

tion. Computers, Environment and Urban Systems, 53, 36–46. https://doi.org/10.1016/

j.compenvurbsys.2014.12.001.

Kemp, S. (2018). Digital in 2018: World's internet users pass the 4 billion mark. Retrieved

from https://wearesocial.com/blog/2018/01/global-digital-report-2018.

Kitchin, R. (2013). Big data and human geography: Opportunities, challenges and risks.

Dialogues in Human Geography, 3(3), 262–267. https://doi.org/10.1177/

2043820613513388.

Laestadius, L. (2017). Instagram. In A. Q.-H. Luke Sloan (Ed.). The SAGE Handbook of

Social Media Research Methods (pp. 573–592). London: SAGE Publications Ltd.

de Lange, M., & de Waal, M. (2013). Owning the city: New media and citizen engagement in

urban design. (First Monday).

Lansley, G., & Longley, P. A. (2016). The geography of Twitter topics in London.

Computers, Environment and Urban Systems, 58, 85–96. https://doi.org/10.1016/j.

compenvurbsys.2016.04.002.

Lee, R., Wakamiya, S., & Sumiya, K. (2013). Urban area characterization based on crowd

behavioral lifelogs over Twitter. Personal and Ubiquitous Computing, 17(4), 605–620.

https://doi.org/10.1007/s00779-012-0510-9.

Leetaru, K., Wang, S., Cao, G., Padmanabhan, A., & Shook, E. (2013). Mapping the global

Twitter heartbeat: The geography of Twitter. First Monday, 18.

Liftn, J., & Parad (2018). Dual reality: Merging the real and Virtual. OpenStreetMap Wiki.

Retrieved from http://wiki.openstreetmap.org/w/index.php?title=Browsing&

oldid=1550720.

Liu, L., Zhou, B., Zhao, J., & Ryan, B. D. (2016). C-IMAGE: City cognitive mapping

through geo-tagged photos. GeoJournal, 81(6), 817–861. https://doi.org/10.1007/

s10708-016-9739-6.

Lloyd, A., & Cheshire, J. (2017). Deriving retail Centre locations and catchments from

geo-tagged Twitter data. Computers, Environment and Urban Systems, 61, 108–118.

https://doi.org/10.1016/j.compenvurbsys.2016.09.006.

López Baeza, J., Serrano Estrada, L., & Nolasco-Cirugeda, A. (2016). Percepción y uso

social de una transformación urbana a través del social media. Las setas gigantes de la

calle San Francisco. I2 Innovación e Investigación En Arquitectura y Territorio.Vol. 4.Las

setas gigantes de la calle San Francisco. I2 Innovación e Investigación En Arquitectura y

Territorio (pp. 2–). . https://doi.org/10.14198/i2.2016.5.03.

Luo, F., Cao, G., Mulligan, K., & Li, X. (2016). Explore spatiotemporal and demographic

characteristics of human mobility via Twitter: A case study of Chicago. Applied

Geography, 70, 11–25. https://doi.org/10.1016/j.apgeog.2016.03.001.

Lynch, K. (1960). The image of the city. MIT Press.

Mahto, D. K., & Singh, L. (2016). A dive into Web Scraper world. 2016 3rd International

Conference on Computing for Sustainable Global Development (INDIACom) (pp. 689–

693). .

Manovich, L. (2016). Notes on Instagrammism and mechanisms of contemporary cultural

identity (and also photography, design, Kinfolk, k- pop, hashtags, mise-en-scène, and

cостояние). Instagram and Contemporary image.

Martí, P., Serrano-Estrada, L., & Nolasco-Cirugeda, A. (2017). Using locative social media

and urban cartographies to identify and locate successful urban plazas. Cities, 64,

66–78. https://doi.org/10.1016/j.cities.2017.02.007.

Marwick, A. E., & Boyd, d. (2011). I tweet honestly, I tweet passionately: Twitter users,

context collapse, and the imagined audience. New Media & Society, 13(1), 114–133.

https://doi.org/10.1177/1461444810365313.

Mayr, P., & Weller, K. (2017). Think before you collect: Setting up a data collection ap-

proach for social media studies. In L. Sloan, & A. Quan-Haase (Eds.). The SAGE

Handbook of Social Media Research Methods (pp. 108–124). London: SAGE

Publications Ltd.

McCarney, R., Warner, J., Iliﬀe, S., van Haselen, R., Griﬃn, M., & Fisher, P. (2007). The

Hawthorne eﬀect: A randomised, controlled trial. BMC Medical Research Methodology,

7(30), https://doi.org/10.1186/1471-2288-7-30.

McLain, R., Poe, M., Biedenweg, K., Cerveny, L., Besser, D., & Blahna, D. (2013). Making

sense of Human Ecology Mapping: An Overview of Approaches to Integrating Socio-

Spatial Data into Environmental Planning. Human Ecology, 41(5), 651–665. https://

doi.org/10.1007/s10745-013-9573-0.

Milne, D., Thomas, P., & Paris, C. (2012). Finding, Weighting and describing Venues :

CSIRO at the 2012 TREC Contextual Suggestion Track. The Twenty-ﬁrst Text REtrieval

Conference (TREC 2012) Proceedings.

Moreno Izquierdo, L., Ramón Rodríguez, A., & Such Devesa, M. J. (2016). Turismo co-

laborativo stá Airbnb transformando el sector del alojamiento? Economistas, 150,

107–119.

Morstatter, F., Pfeﬀer, J., Liu, H., & Carley, K. M. (2013). Is the Sample good enough?

Comparing Data from Twitter's Streaming API with Twitter's Firehose. 400–408. https://

doi.org/10.1007/978-3-319-05579-4_10.

Noulas, A., Scellato, S., Mascolo, C., & Pontil, M. (2010). An Empirical Study of

Geographic User activity patterns in Foursquare. Fifth International AAAI Conference

on Weblogs and Social Media (pp. 570–573). .

Noulas, A., Scellato, S., Lambiotte, R., Pontil, M., & Mascolo, C. (2012). A tale of many

cities: Universal patterns in human urban mobility. PLoS One, 7(5), https://doi.org/

10.1371/journal.pone.0037027.

Panteras, G., Wise, S., Lu, X., Croitoru, A., Crooks, A., & Stefanidis, A. (2015).

Triangulating Social Multimedia Content for Event Localization using Flickr and

Twitter. Transactions in GIS, 19(5), 694–715. https://doi.org/10.1111/tgis.12122.

Peña-López, I., Congosto, M., & Aragón, P. (2014). Spanish Indignados and the Evolution

of the 15M Movement on Twitter: Towards Networked Para-institutions. Journal of

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

Spanish Cultural Studies, 1–28.

Pew Research Center (2017). Social media fact sheet. Retrieved 15 May 2016, from

http://www.pewinternet.org/fact-sheet/social-media/.

Quercia, D. (2015). Chatty, Happy, and Smelly Maps. Proceedings of the 24th International

Conference on World Wide Web, 741. https://doi.org/10.1145/2740908.2741717.

Quercia, D. (2016). Playful Cities : Crowdsourcing Urban Happiness with Web Games. 42, 3.

Quercia, D., & Saez, D. (2014). Mining urban deprivation from Foursquare: Implicit

crowdsourcing of city land use. IEEE Pervasive Computing, 13(2), 30–36. https://doi.

org/10.1109/MPRV.2014.31.

Quercia, D., Aiello, L. M., Mclean, K., & Schifanella, R. (2015a). Smelly Maps: The Digital

Life of Urban Smellscapes. AAAI Publications327–336.

Quercia, D., Aiello, L. M., Schifanella, R., & Davies, A. (2015b). The Digital Life of Walkable

Streets. 875–884. https://doi.org/10.1145/2736277.2741631.

Roberts, H. V. (2017). Using Twitter data in urban green space research: A case study and

critical evaluation. Applied Geography, 81, 13–20. https://doi.org/10.1016/j.apgeog.

2017.02.008.

Roick, O., & Heuser, S. (2013). Location based social networks–deﬁnition, current state of

the art and research agenda. Transactions in GIS, 17(5), 763–784.

Saker, M., & Evans, L. (2016). Locative Media and Identity: Accumulative Technologies of

the Self. SAGE Open, 6(3), https://doi.org/10.1177/2158244016662692.

Samet, H. (1984). The Quadtree and Related Hierarchical Data Structures. ACM

Computing Surveys, 16(2), 187–260. https://doi.org/10.1145/356924.356930.

Sang, E. T. K., & Van Den Bosch, A. (2013). Dealing with big data: The case of Twitter.

Computational Linguistics in the Netherlands Journal, 3, 121–134. https://doi.org/10.

1126/science.345.6193.148-a.

Serrano-Estrada, L., Marti, P., & Nolasco-Cirugeda, A. (2016). Comparing two Residential

Suburban areas in the Costa Blanca, Spain, Articulo. Journal of Urban Research, 13.

https://doi.org/10.4000/articulo.2935.

Shelton, T., Poorthuis, A., & Zook, M. (2015). Social media and the city: Rethinking urban

socio-spatial inequality using user-generated geographic information. Landscape and

Urban Planning, 142, 198–211. https://doi.org/10.1016/j.landurbplan.2015.02.020.

Silva, T. H., Vaz De Melo, P. O. S., Almeida, J. M., Salles, J., & Loureiro, A. A. F. (2014).

Revealing the City that we cannot see. ACM Transactions on Internet Technology

(TOIT), 14(4), 26.

Sloan, L. (2017). Social Science ‘Lite’? Deriving Demographic Proxies from Twitter. In L.

Sloan, & A. Quan-Haase (Eds.). The SAGE Handbook of Social Media Research Methods

(pp. 90–104). London: SAGE Publications Ltd.

Sloan, L., & Morgan, J. (2015). Who tweets with their location? Understanding the re-

lationship between demographic characteristics and the use of geoservices and geo-

tagging on twitter. PLoS One, 10(11), 1–15. https://doi.org/10.1371/journal.pone.

0142209.

Sloan, L., & Quan-Haase, A. (2017). The SAGE Handbook of Social Media Research Methods.

London: SAGE Publications Ltd. Retrieved from https://www.amazon.es/Handbook-

Social-Media-Research-Methods/dp/1473916321.

Soja, E. (1989). Postmodern geographies. The reassertion of space in critical social theory.

London, New York: Verso.

Steiger, E., Westerholt, R., Resch, B., & Zipf, A. (2015). Twitter as an indicator for

whereabouts of people? Correlating Twitter with UK census data. Computers,

Environment and Urban Systems, Vol. 54, 255–265. https://doi.org/10.1016/j.

compenvurbsys.2015.09.007.

Sui, D., & Goodchild, M. (2011). The convergence of GIS and social media: Challenges for

GIScience. International Journal of Geographical Information Science, 25(11),

1737–1748. https://doi.org/10.1080/13658816.2011.604636.

Tasse, D., & Hong, J. I. (2014). Using social media data to understand cities. NSC work-

shops on big data and urban informatics, Chicago.

Temes Cordóvez, R. R., Simancas Cruz, M. R., Peñarrubia Zaragoza, M. P., Moya Fuero,

A., & García Amaya, A. M. (2016). Characterization and spatial identiﬁcation of

holiday tourist assessments in the city of Valencia. In J. Rivas Navarro, & B. Bravo

Rodríguez (Eds.). 6th Sustainable Development Symposium - Book of Abstracts. Granada:

Godei.

Tsou, M. H., Yang, J. A., Lusher, D., Han, S., Spitzberg, B., Gawron, J. M., & An, L. (2013).

Mapping social activities and concepts with social media (Twitter) and web search

engines (Yahoo and Bing): A case study in 2012 US Presidential Election. Cartography

and Geographic Information Science, 40(4), 337–348. https://doi.org/10.1080/

15230406.2013.799738.

Tufekci, Z. (2014). Big questions for social media big data: Representativeness, validity

and other methodological pitfalls. ICWSM ‘14: Proceedings of the 8th International

AAAI Conference on Weblogs and Social Media (pp. 505–514). .

Twitter, I. (2018a). Rate limiting. Retrieved from https://developer.twitter.com/en/

docs/basics/rate-limiting.html.

Twitter, I. (2018b). Tweet location FAQs. Retrieved 14 May 2018, from https://help.

twitter.com/en/safety-and-security/tweet-location-settings.

Van Canneyt, S., Schockaert, S., Van Laere, O., & Dhoedt, B. (2012a). Detecting places of

interest using social media. Proceedings - 2012 IEEE/WIC/ACM International

Conference on Web Intelligence, WI 2012 (pp. 447–451). . https://doi.org/10.1109/WI-

IAT.2012.19.

Van Canneyt, S., Van Laere, O., Schockaert, S., & Dhoedt, B. (2012b). Using social media

to ﬁnd places of interest. Proceedings of the 1st ACM SIGSPATIAL International

Workshop on Crowdsourced and Volunteered Geographic Information - GEOCROWD

‘12https://doi.org/10.1145/2442952.2442954.

Villatoro, D., Serna, J., Rodríguez, V., & Torrent-Moreno, M. (2013). The TweetBeat of the

City: Microblogging used for Discovering Behavioural patterns during the MWC2012

BT. In J. Nin, & D. Villatoro (Vol. Eds.), Citizen in Sensor Networks. Lecture Notes in

Computer Science.Vol. 7685.Citizen in Sensor Networks. Lecture Notes in Computer

Science (pp. 43–56). Berlin, Heidelberg: Springer Berlin Heidelberg. https://doi.org/

10.1007/978-3-642-36074-9_5.

Wang, W. (2013). Using Location-based Social Media for Ranking Individual Familiarity

wih Places: A Case Study with Foursquare Check-in Data. In G. Gartner, & H. Huang

(Eds.). Progress in Location- Based Services 2014 (pp. 171–183). Springer.

Wilken, R. (2014). Places nearby : Facebook as a location-based social media platform.

New Media & Society, 16(7), 1087–1103. https://doi.org/10.1177/

1461444814543997.

Williams, S. (2012). We are here now. Social media and the psychological city. Retrieved

from http://weareherenow.org/about.html.

Williams, M. J., & Chorley, M. J. (2017). Foursquare. In L. Sloan, & A. Quan-Haase (Eds.).

The SAGE Handbook of Social Media Research Methods (pp. 610–626). London: SAGE

Publications Ltd.

Yang, L., Sun, T., Zhang, M., & Mei, Q. (2012). We know what@ you# tag: Does the dual

role aﬀect hashtag adoption? WWW’12 Proceedings of the 21st International Conference

on World Wide Web (pp. 261–270). . https://doi.org/10.1145/2187836.2187872.

P. Martí et al. Computers, Environment and Urban Systems xxx (xxxx) xxx–xxx

Exploring city dynamics through tweets: A framework for capturing urban activities as complex spatiotemporal patterns

Article

Full-text available

Apr 2025
CITIES

This paper presents a novel framework for analysing urban activities as spatiotemporal patterns using Location-Based Social Media (LBSM) data. The methodology integrates the spatial, temporal, and semantic dimensions of geolocated tweets to investigate cities as Complex Adaptive Systems (CAS) and their relationship with urban form. By combining spatiotemporal clustering (ST-DBSCAN) and topic modelling (LDA), the framework uncovers dynamic activity patterns shaped by top-down mechanisms and bottom-up self-organizing behaviours. A custom tool and Graphical User Interface was developed to support data exploration and experimentation, enabling the contextual analysis of activity clusters. The framework was tested in Manchester City Centre as an exploratory case study, focusing on the impact of Covid-19 lockdown measures as a significant disturbance. The results reveal how urban characteristics, urban form, and social behaviours influence activity levels and patterns, demonstrating fluctuations that highlight different degrees of adaptability. By exploring cities as hybrid urban-digital spaces, this approach provides an alternative approach for understanding cities as CAS, linking space to place and for exploring adaptive behaviour. The paper concludes by reflecting on the framework, use of LBSM for researching cities, and outlining directions for future work of comparing cities and integrating alternative data.

Decoding Public Sentiment Topics in Google Map Reviews on Urban Infrastructure Development of Belt and Road Initiative

Preprint

Full-text available

Feb 2025

Despite the broad scope of Belt and Road Initiative (BRI) projects and deep intertwine with urban development, there is a lack of quantitative research utilizing crowd-sourced data to understand public perceptions, particularly from both spatial and temporal perspectives. This study analyzes 144,210 Google reviews from 352 BRI urban infrastructure projects between 2012 and 2023, encompassing six urban infrastructure categories. Using the Valence Aware Dictionary and Sentiment Reasoner for sentiment analysis and Multi-grained Latent Dirichlet Allocation for topic modeling, the study reveals that sentiment for BRI projects is generally positive, especially in upper-middle-income countries. Discussion topics can be clustered into professional function (44%), benefits/disbenefits (24%), service industry (19%), and development (13%). Higher-income areas focus on service-related topics, while lower-income areas emphasize development. Moreover, higher urban growth rates at country level correlate with more positive sentiments and a greater focus on development. However, high investment areas experience more polarized reviews, indicating unmet expectations. Besides, the urbanization process at city level and local level also impact the performance of BRI projects, suggesting the importance of integrating BRI projects with local community. This study contributes to the understanding of the complex interplay between BRI projects, urban development, and public perception across regions and over time.

Mapeando la ciudad a través del BigData. Análisis temporal de las dinámicas urbanas

Conference Paper

Oct 2024

En la literatura se ha recogido extensamente la relación existente entre el comportamiento social y la forma de la ciudad, entendida no solo desde el punto de vista morfológico, sino también funcional. No obstante, en general, esta relación se ha analizado de forma sectorial, en temáticas muy concretas como la movilidad, la turistificación o las dinámicas comerciales. En los últimos años, estos estudios han evolucionado por la disponibilidad de datos provenientes de la digitalización de los sistemas o las redes sociales que permiten realizar una aproximación más real y masiva al comportamiento de las personas. En ese sentido, los datos masivos georeferenciados constituyen una oportunidad para estudiar fenómenos urbanos y mapear las dinámicas que se generan en las ciudades. En este contexto se ha desarrollado el proyecto DINUR, Método de análisis de las dinámicas urbanas a través de Big(Geo)Data para la regeneración y transformación de la ciudad, financiado por la Diputación de Gipuzkoa cuyo objetivo ha sido desarrollar un método para la integración, visualización y análisis exploratorio de datos sociales urbanos heterogéneos y a gran escala que facilite la comprensión de las dinámicas urbanas a escala municipal. Se ha aplicado al caso estudio de Donostia-San Sebastián. Una vez establecidos los 3 bloques de trabajo (espacios estanciales, hitos y flujos), la parte sustancial del proyecto ha consistido en analizar las fuentes y validar la información disponible. A partir de ello, se han definido los indicadores y el método de cálculo y visualización para cada caso. La aplicación en el caso de Donostia-San Sebastián ha permitido realizar una aproximación espacio-temporal del uso que hacen las personas de la ciudad y un análisis secuencial de las dinámicas urbanas para cada bloque de trabajo. En esta comunicación se muestran los resultados obtenidos en la visualización de los mismos y las aportaciones más relevantes de la investigación donde la cartografía urbana trasciende de lo geográfico y da un salto a lo social a través de las interaccciones de las personas, recontextualizando la ciudad.

From Policy to Platforms: Analysing Public Engagement with Singapore's Smart Nation Initiative through Social Media Discourse

Article

Full-text available

May 2025

Reza Shaker

LATA: A Pilot Study on LLM-Assisted Thematic Analysis of Online Social Network Data Generation Experiences

Article

Full-text available

May 2025

Large Language Models (LLMs) have gained attention in research and industry, aiming to streamline processes and enhance text analysis performance. Thematic Analysis (TA), a prevalent qualitative method for analyzing interview content, often requires at least two human experts to review and analyze data. This study demonstrates the feasibility of LLM-Assisted Thematic Analysis (LATA) using GPT-4 and Gemini. Specifically, we conducted semi-structured interviews with 14 researchers to gather insights on their experiences generating and analyzing Online Social Network (OSN) communications datasets. Following Braun and Clarke's six-phase TA framework with an inductive approach, we initially analyzed our interview transcripts with human experts. Subsequently, we iteratively designed prompts to guide LLMs through a similar process. We compare and discuss the manually analyzed outcomes with responses generated by LLMs and achieve a cosine similarity score up to 0.76, demonstrating a promising prospect for LATA. Additionally, the study delves into researchers' experiences navigating the complexities of collecting and analyzing OSN data, offering recommendations for future research and application designers.

DESIGNING AND IMPLEMENTING A MULTI-CRITERIA INTELLIGENT FRAMEWORK TO ENSURE THE RELEVANCE OF REQUIREMENTS SOURCED FROM SOCIAL NETWORK SITES

Article

Full-text available

Apr 2025

The proliferation of social networking sites (SNS) and software failures primarily arising from the requirements elicitation phase, motivated researchers to develop methods, that incorporate SNS-based users’ needs into the requirements engineering stage, crucial for developing reliable software. This approach improves user-centric needs and identifies innovative features, but relevance verification has not been thoroughly examined, leading to challenges in filtering and prioritizing relevant information emanated from jargon, informal language, and diverse expressions detected in user comments. This research proposes a novel intelligent framework for relevance verification of SNS-sourced requirements, combining multiple criteria like organizational goals, business rules, related service datasets, and user comments. The proposed framework balances user, organization, and developer needs by simplifying the process of determining relevant requirements and enhancing their validity. The framework learns and utilizes consolidated criteria features as regulation mechanisms, addressing challenges in isolating relevant users' needs and minimizing traditional method limitations. The study uses qualitative methods for framework development and empirical research methods for sentiment and trend analysis, combining customized word embedding models and natural language processing. A case study of digital healthcare systems in developing nations explores three evaluation categories: business rules, related service datasets, and a blend of both. The dataset includes 2400 key phrases from 800 user needs, 540 business goals, and 900 from service datasets.The proposed method achieved a relevance rate of 88%, surpassing individual methods. The study contributes to IT and software engineering fields by providing a novel framework for relevance verification of SNS-based requirements, ensuring their alignment with actual user needs and improving their completeness and prioritization, leading to significant enhancements in system design and development.

Exploring the intersection of cyberspace and urban geography: The influence of instagram's #Bookgarden on the popularity dynamics of Tehran's urban spaces

Article

Full-text available

Apr 2025

Background With advancements in new technologies, unstructured data can be extracted from the virtual world. Identifying the relationship between cyberspace and real space in order to evaluate urban spaces is valuable. Instagram social media, with its facilities, is in the category of the mentioned data sources. Users can impact a specific place in real space. Accordingly, the purpose of this study was to evaluate and measure the popularity of the urban space of Tehran Book Garden based on the #book garden and the location of active users on Instagram. The question is the impact of the #book garden on the popularity of this urban space. Methods First, the required data on Instagram were extracted through the application programming interface (API) and analyzed through networking on geographical maps. Next, by discovering that a certain part of the urban area has irregular networking based on the #book garden, the most active area was measured. Results and Conclusions Finally, with interpolation calculation, a zoning map of the #book garden effectiveness was produced. The location of the Book Garden is an effective area and cultural, scientific, and artistic area, which is introduced as a hidden layer. This study developed a new, innovative, and reliable method for measuring the attractiveness of urban spaces that can be used as a pattern in other studies.

AI-Driven Social Media Analysis for Disaster Situational Awareness in Smart Urban Environments

Chapter

Jan 2025

This paper explores the transformative impact of social media analytics in disaster response. By utilizing advanced algorithms and machine learning, social media analytics enables the rapid processing of vast data streams, delivering real-time insights and trends crucial for effective disaster management. The paper highlights how social media has evolved into a vital tool for crisis communication, geolocation-based disaster response, and sentiment analysis, offering valuable insights into the emotional and psychological effects of disasters. It addresses the challenges of integrating AI in this realm, emphasizing the necessity of data privacy, ethical considerations, and transparent public engagement. The paper shows both the huge potential and inherent difficulties of using AI-driven social media analysis in disaster management by looking at a wide range of real-life case studies and opportunities within smart urban environments.

Architecture in the age of social media: introduction to the special issue

Article

Apr 2025
ArchNet-IJAR

Purpose This special issue brings together the work of scholars investigating architecture’s interaction with social media across a broad spectrum of geographical contexts, methodological approaches and theoretical frameworks. The aim of this effort is twofold: to provide a snapshot of the current state of the field of research concerned with architecture as it is influenced by the unique features of social media and to delineate something that might become a shared agenda through which the field might be advanced. Design/methodology/approach The introduction provides a critical overview of the papers in the special issue and how they connect to the field. Findings The entanglements of architecture with social media are complex, multi-valent and heterogeneous, and an understanding of architecture’s relationship with social media is still unfolding. Future scholarship will require interdisciplinary approaches, drawing on methods from fields such as media studies, sociology, anthropology, visual culture and data science. While much of the existing scholarship focuses on case studies – individual buildings, firms or cities – there is a need for more systematic research that can address broader questions. Originality/value This collection maps a variable terrain rather than proposing a singular path forward for scholarship. Architecture continues to respond to the logics of media, even as media becomes social in new and unexpected ways. The importance of studying how architecture is affected by social media extends beyond the profession itself: the actions people take online end up having impacts on buildings and urban spaces as well as social realities.

Disseminating Sustainable End-of-life Product Disposal Information Through Social Media Nudging

Chapter

Apr 2025

The invention of single-use baby diapers came as a relief to most mothers who bore the burden of baby caregiving. The common practice is to dispose of both the diapers and the waste matter. However, this practice has been linked to various forms of contamination ranging from air, land, and water pollution. As a way of promoting responsible and sustainable discarding of single-use baby diapers, a social media skit was developed to educate baby caregiving women on appropriate disposal methods. The intervention took an ecofeminism perspective since women are traditionally socialized to be care givers and generally suffer the brunt of environmental degradation. The primary objective was to determine the efficacy of social media skits in bringing awareness as well as achieving attitudinal and behavioral changes with regard to proper separation of waste material at source. A quasi-experimental design was adopted that involved exposing 100 baby caregiving mothers using single-use baby diapers in an experimental group to an educative social media skit and then assess their levels of environmental awareness, attitudes as well as waste disposal practices. The same outcomes were also assessed on the control group comprising 100 women. Data was analyzed using one-way multivariate analysis of variance. The results indicated that social media skits have a statistically significant effect on environmental awareness, attitudes, and waste disposal practices. This chapter recommends the development and dissemination of social media skits through several online platforms that are accessible to most baby caregiving mothers.

Geolocated social media as a rapid indicator of park visitation and equitable park access

Article

Full-text available

Feb 2018
COMPUT ENVIRON URBAN

Understanding why some parks are used more regularly or intensely than others can inform ways in which urban parkland is developed and managed to meet the needs of a rapidly expanding urban population. Although geolocated social media (GSM) indicators have been used to examine park visitation rates, studies applying this approach are generally limited to flagship parks, national parks, or a small subset of urban parks. Here, we use geolocated Flickr and Twitter data to explore variation in use across New York City's 2143 diverse parks and model visitation based on spatially-explicit park characteristics and facilities, neighborhood-level accessibility features and neighborhood-level demographics. Findings indicate that social media activity in parks is positively correlated with proximity to public transportation and bike routes, as well as particular park characteristics such as water bodies, athletic facilities, and impervious surfaces, but negatively associated with green space and increased proportion of minority ethnicity and minority race in neighborhoods in which parks are located. Contrary to previous studies which describe park visitation as a form of nature-based recreation, our findings indicate that the kinds of green spaces present in many parks may not motivate visitation. From a social equity perspective, our findings may imply that parks in high-minority neighborhoods are not as accessible, do not accommodate as many visitors, and/or are of lower quality than those in low-minority neighborhoods. These implications are consistent with previous studies showing that minority populations disproportionately experience barriers to park access. In applying GSM data to questions of park access, we demonstrate a rapid, big data approach for providing information crucial for park management in a way that is less resource-intensive than field surveys.

Quantitative Comparison of Open-Source Data for Fine-Grain Mapping of Land Use

Conference Paper

Full-text available

Nov 2017

This paper performs a quantitative comparison of open-source data available on the Internet for the fine-grain mapping of land use. Three points of interest (POI) data sources--Google Places, Bing Maps, and the Yellow Pages--and one volunteered geographic information data source--Open Street Map (OSM)--are compared with each other at the parcel level for San Francisco with respect to a proposed fine-grain land-use taxonomy. The sources are also compared to coarse-grain authoritative data which we consider to be the ground truth. Results show limited agreement among the data sources as well as limited accuracy with respect to the authoritative data even at coarse class granularity. We conclude that POI and OSM data do not appear to be sufficient alone for fine-grain land-use mapping.

Mapping Socials: A Voluntary Map of a Great Event in Monza Park

Article

Full-text available

Nov 2017

The paper concerns a study developed on the largest enclosed park in Europe, in the town of Monza, perceived and returned in a series of icon-maps through the use of social twitter and instagram in a specific time frame. The aim is to investigate the contribution provided by social data through mobile phone network considered as a tool to describe a temporary event, and to describe the shape of a town that does not physically exist but can be drowned always in a different way depending on several variable data.

Using locative social media and urban cartographies to identify and locate successful urban plazas

Article

Full-text available

Feb 2017
CITIES

Locative social media networks as open sources of data allow researchers and professionals to acknowledge which city places are preferred, used and livable. Following this hypothesis, this paper proposes a methodology to identify successful public spaces – plazas – through the location-based social media network Foursquare and to analyze their urban position using morphological and historical cartographies. The overall methodology comprises three stages. First, the most important cities of the province of Alicante were selected. Second, the most relevant plaza of each city was identified using data retrieved from the social network Foursquare. Finally, the location of each plaza is analyzed in relation to the historic center and the main axes of the city. Possible correlations between their urban location and their vibrant character were subsequently identified. Two findings have emerged from this study: (a) a strong spatial relationship exists between the most successful plazas and the historic city center, which reinforces their traditional social character; and (b) all plazas share two similar traits, their location within the urban network and their proximity to the main axes of the city.

Using Twitter data in urban green space research: A case study and critical evaluation

Article

Full-text available

Apr 2017
APPL GEOGR

Helen Roberts

Percepción y uso social de una transformación urbana a través del social media. Las setas gigantes de la calle San Francisco

Article

Full-text available

Dec 2016

El análisis del espacio público urbano a partir de su dimensión perceptual ha sido, tradicionalmente, un recurso fundamental para reconocer qué características físicas lo diferencian y le confieren actividad y vitalidad. Esta aproximación es particularmente relevante en entornos urbanos transformados. Conocer la percepción social sobre el espacio público permite valorar en qué medida las acciones realizadas modifican las dinámicas urbanas. Bajo esta consideración, el objetivo de este trabajo es estudiar la vida urbana tras la transformación de la Calle San Francisco, en Alicante, España, en un espacio lúdico caracterizado por la implantación de una serie de setas gigantes. Para ello, se definieron y evaluaron tres indicadores de vida pública: las actividades económicas en la edificación antes y después de la intervención; el uso social del espacio proyectado; y, la imagen pública —o imagen percibida—resultante. Se propone una metodología híbrida que combina los datos recogidos en un estudio de campo con los datos provenientes de fuentes de base tecnológica, concretamente, de la red social Instagram y del servicio web Google Street View. Los resultados obtenidos ponen de manifiesto que la utilización de datos offline y online permite confirmar, complementar o cuestionar las deducciones parciales de cada uno de los métodos, ratificando así que la transformación de la Calle San Francisco ha modificado sustancialmente no sólo su identidad física, sino sobre todo la imagen colectiva del espacio con relación a la del resto de la ciudad. En definitiva, este trabajo evidencia la pertinencia de incorporar técnicas de base tecnológica para entender la complejidad de las relaciones sociales, físicas y virtuales, que se producen como consecuencia de la transformación de un espacio urbano.

The SAGE Handbook of Social Media Research Methods

Book

Jan 2016

Discovery of spatio temporal patterns from location-based social networks

Article

Mar 2016

Location-based social networks (LBSNs) such as Twitter or Instagram are a good source for user spatiooral behaviour. These networks collect data from users in such a way that they can be seen as a set of collective and distributed sensors of a geographical area. A low rate sampling of user's location information can be obtained during large intervals of time that can be used to discover complex patterns, including mobility profiles, points of interest or unusual events. These patterns can be used as the elements of a knowledge base for different applications in different domains such as mobility route planning, touristic recommendation systems or city planning. The aim of this paper is twofold, first to analyse the frequent spatiooral patterns that users share when living and visiting a city. This behaviour is studied by means of frequent itemsets algorithms in order to establish some associations among visits that can be interpreted as interesting routes or spatiooral connections. Second, to analyse how the spatiooral behaviour of a large number of users can be segmented in different profiles. These behavioural profiles are obtained by means of clustering algorithms that show the different patterns of behaviour of visitors and citizens. The data analysed were obtained from the public data feeds of Twitter and Instagram within an area surrounding the cities of Barcelona and Milan for a period of several months. The analysis of these data shows that these kinds of algorithms can be successfully applied to data from any city (or general area) to discover useful patterns that can be interpreted on terms of singular places and areas and their temporal relationships.

City dynamics through Twitter: Relationships between land use and spatiotemporal demographics

Article

Feb 2018
CITIES

Social network data offer interesting opportunities in urban studies. In this study, we used Twitter data to analyse city dynamics over the course of the day. Users of this social network were grouped according to city zone and time slot in order to analyse the daily dynamics of the city and the relationship between this and land use. First, daytime activity in each zone was compared with activity at night in order to determine which zones showed increased activity in each of the time slots. Then, typical Twitter activity profiles were obtained based on the predominant land use in each zone, indicating how land uses linked to activities were activated during the day, but at different rates depending on the type of land use. Lastly, a multiple regression analysis was performed to determine the influence of the different land uses on each of the major time slots (morning, afternoon, evening and night) through their changing coefficients. Activity tended to decrease throughout the day for most land uses (e.g. offices, education, health and transport), but remained constant in parks and increased in retail and residential zones. Our results show that social network data can be used to improve our understanding of the link between land use and urban dynamics.

An empirical study of geographic user activity patterns in foursquare

Article