Profiles of mapping of content creators in a geo-social network: The case of 21 Brazilian cities
First Monday

Profiles of mapping of content creators in a geo-social network: The case of 21 Brazilian cities by Aline Marques Morais and Nazareno Andrade

A geo-social network (GSN) is a type of social network where geolocated information shared among users plays an important role. The fast development of GSNs in recent years has made this system a valuable tool for people to learn about places on a large scale, using information provided by non-specialists. This paper examines how value is created in GSNs by studying the collective behavior of content creators in a popular GSN, Foursquare. In particular, this study examines the multifaceted aspect of content contribution behavior, and investigates how it occurs in multiple cities of an area previously studied by few: 21 medium- and large-sized Brazilian cities. Our study leverages cluster analyses to explain how users differ in how they collaborate. We discovered four profiles of content creators, and identify how such profiles are useful to GSNs.


Related work
Data and methods




Geo-social networks (GSNs) are collaborative systems with a main component involving geolocated information. Such systems combine online social networks and geolocated information to contextualize content in a new form of online collaboration. This new configuration of content helps analysts to map urban areas (Liang, et al., 2013) and end users decide about places in a specific area (Gao and Liu, 2014). These systems have been gaining significant attention, largely due to the increasing use of mobile devices. From 2014 to 2020, the number of smartphone users worldwide is expected to grow to 1.3 billion (Statista, 2014).

In GSNs, users annotate space and interact to disclose useful information. As in other collaborative systems, value created by a GSN depends primarily on users acting as content creators. The popularization of GSNs has stimulated studies on these collaborators (Abraham, et al., 2009). These studies have concluded that GSN content creators are heterogeneous according to motivations (Noulas, et al., 2011), habits (Morais and Andrade, 2014) and preferences (Chen, et al., 2016). However, the diversity of creators in GSNs has received little attention (Yang, et al., 2016; Wu and Li, 2016). So, understanding this diversity and how it is manifested is paramount to system designers, administrators and users alike.

This work focuses on GSN content creators in urban areas. We are motivated by the scarcity of experiments about GSN creators in multiple cities. We detected a gap in the quantitative disparity of studies of GSN users in urban areas in different parts of the world. For example, cities in U.S. (Noulas, et al., 2011; Chang and Sun, 2011; Gao, et al., 2015) and Europe (Yang, et al., 2016; Noulas, et al., 2011) have received considerable attention in the analysis of GSNs in contrast to cities in Brazil (Morais and Andrade, 2014), Canada, Mexico and China (Hjorth, et al., 2012).

This study is centered around GSN content creators who collaborated in urban areas of Brazil. This country is the fifth largest growing market of new smartphones in the world (Statista, 2014), but has received little attention from GSN studies so far. Besides, this work focuses on GSNs that were considered useful during decision-making processes of choosing places to visit. The GSN Foursquare is a prototypical example of such a system. In Foursquare, creators annotate venues of a city where they make check-ins and create hints that leave geolocated suggestions and comments for other users. Presently, Foursquare has 50 million active users and more than 75 million shared tips (Weber and Novet, 2015). Regarding Foursquare acceptation, Brazil is the third largest public of Foursquare, behind U.S. and Indonesia.

This paper leverages cluster algorithms to segment content creators into four profiles that reveal the typical combinations of activity distribution over time, space and characteristics of venues in the GSN. Moreover, studying multiple cities allows us to compare creator profiles per urban area. We collected and analyzed data of Foursquare users in 21 large Brazilian cities.

Regarding the research methodology of studies of GSN creators, the main difficulty of analysis is related to big datasets and correct portraits. In our experiment, the main limitation is the correct description of Foursquare creators, in face of reduced feedback, compared with other GSNs, like Instagram and Facebook Places. The solution was overcome by a long-time observation of interactions in cities (2013–2017). More details on this will be given later in this paper.

The remainder of this paper is organized as follows. The next section discusses the literature of user diversity on GSNs. Afterward, we present metrics about content creators used in the cluster analysis and present data and methods of our observational study. The last two sections describe results about the diversity of content creators and discuss that diversity in a larger context.



Related work

The main value of a GSN is its peer-production, where users act as content creators. As in other systems, understanding how these content creators behave is paramount to understanding how to best design and operate GSNs. There are several ways to characterize content creators in GSNs. This work considers dimensions which stem from an analysis of factors deemed relevant in related work on GSNs:

  1. Background features of creators: Previous work has considered variables related to demographic details of creators such as gender (Zhang, et al., 2016), age and hometown (Kang and Lerman, 2011);

  2. Features of content shared by creators: Another body of work focuses on variables that provide feedback on shared content in GSNs. These studies have analyzed the influence of creators according to relevance (Morais and Andrade, 2014) and the utility of creators (Schaar, et al., 2013) for other GSN users. An influent creator is someone who provides geolocated information with significant effects for other users. This mapping is useful for setting preferences and patterns for creators;

  3. Features of the collaborative behavior of creators: The variables in this group measure features such as the number of interactions performed in a given time span, including posts, check-ins and comments. They indicate the intensity of collaboration in specific places (Yang, et al., 2016; Cranshaw, et al., 2012) and groups (Jin, et al., 2016; Brown, et al., 2014), or the variety implied in the sum of distinct places where creators have collaborated. Studies describe preferences of creators (Chorley, et al., 2015) in order to understand the distribution of diversity by Shannon’s index (Straathof, 2007), leveraged to study the dynamics of cities (Silva, et al., 2013a). Moving from space to time distribution of collaboration activities, studies have used the periodicity of contributions to characterize content creators. This metric is useful for capturing aspects of human mobility, using GPS position traces (Preoţiuc-Pietro and Cohn, 2013), and it is present in mining users (Preoţiuc-Pietro and Cohn, 2013) and definition of parameters for recommendations in GSNs (Rahimi and Wang, 2013);

  4. Spatial features in the behavior of creators: These variables evaluate preferences and habits of creators in urban space. One metric is a variable that measures the distance walked by collaborators during GSN collaboration, which enables mapping routes (Hawelka, et al., 2014) while identifying POIs (points of interest) (Silva, et al., 2013b). It represents social distance, contiguity in terms of geography (Silva, et al., 2014) and social relationships (Wu, et al., 2015; Sarkar, et al., 2016). Another metric is popularity, which characterizes the history of collaborations of GSN creators in different places in a city. Some studies adopt it to assist advertisers in selecting specific sites for effective advertisement placement, or venue owners for improving attractions to customers (Li, et al., 2012). Others works use popularity for the construction of POI networks (Chen, et al., 2015; Gao, et al., 2015). According to Li, et al. (2012), GSN features — place profile, place category, and place age — affect the popularity of a place. The last variable represents the evaluation of places, based on metrics such as the number of collaborations and feedback of users, among others. Normally, an evaluation is represented by a score (Wörndl, et al., 2017) or an index. Metrics of evaluation of places are worthwhile for mapping preferences and patterns.

Most earlier studies considered variables individually or in sets of two (Abraham, et al., 2009; Noulas, et al., 2011; Wakamiya, et al., 2012). Our approach complements these past efforts by considering a multivariate characterization of content creators that enriches our notions of users.

Many earlier experiments often focused on one city (Yu, et al., 2017; Hardy and Lindtner, 2017), while some have studied multiple urban areas (Noulas, et al., 2011; Gao, et al., 2015). Most experiments focus on a single area study in Europe (Agryzkov, et al., 2017; Yang, et al., 2016; Noulas, et al., 2011) and U.S. (Noulas, et al., 2011; Chang and Sun, 2011; Gao, et al., 2015). Others consider areas in Asia (Hjorth, et al., 2012), Africa and South America (Morais and Andrade, 2014; Morais, 2015).

Another set of studies compare cities in different countries (Noulas, et al., 2011; Bawa-Cavia, 2011) or the same country (Gao, et al., 2015). Noulas, et al. (2011) observed GSN users in U.S. cities and London, compared them according to socio-spatial properties, and proposed a novel approach for modeling human activity through place classification. However, it evaluated only the creation of check-ins, not focusing on the creation of reviews or tips about urban spaces. Our study considers a larger and different set of cities, while at the same time characterizes the production of tips, which register more information related to personal impressions, feedback and information deemed useful to other users by a given content creator.



Data and methods

This work is an exploratory study of profiles of GSN content creators in 21 Brazilian cities. The discovery of profiles was done through a partitioning cluster analysis that helped to identify groups of content creators who acted alike. Partitioning cluster algorithms decomposed the sample into a set of disjoint clusters so that each cluster was composed of similar users, while users across clusters were less similar. This study used the k-means cluster algorithm due to its scalability and versatility.

Data collection involved two phases, both using PHP scripts. Firstly, we used the Twitter API to fetch public check-ins of Foursquare in 21 of the largest Brazilian cities from January of 2013. The result of this phase was a list of places visited by the GSN as output data. Secondly, we compiled a set of places from this corpus of check-ins and compiled the tip registers shared in each place through the Foursquare API until May of 2017. Together, the data from places, tips and user names enabled our characterization of content creators.

There were biases in the selected dataset, as noted in other investigations using big data (Honig and MacDowall, 2017; boyd and Crawford, 2012). The first point is about the privacy of users’ information. This work examined details about shared content and collaborative actions of users in the GSN, not personal information. On the quality of the dataset, we mitigated the risk of unreliable information adopting the Twitter API and Foursquare API to secure patterned and authoritative first-hand information. Although Twitter API and Foursquare API presented limits in requisitions per query, this experiment guaranteed the maximum quantity of users and places due to continued observation over an extended period (2013–2017).

The dataset contains 13,783 Foursquare content creators who authored 266,678 tips in the 21 most populous Brazilian cities. These content creators created at least three tips from 24 November 2009 to 22 August 2016. The tip registers present interactions before 2013 because the history per place can contain tips before the date of check-in collecting. Our analysis considered creators with activity 10 times more or less than average as outliers, and as such were not included in this study. The incidence of content creators varied from city to city, according to the proportion of selected users per 100,000 inhabitants of a city, as shown in Figure 1. The proportion of content creators per 100,000 citizens was highest in Recife (114.26 creators per 100,000 inhabitants) and lowest in Salvador (0.93 creators per 100,000 inhabitants).


Proportion of GSN content creators per Brazilian city
Figure 1: Proportion of GSN content creators per Brazilian city.


Metrics for cluster analysis

Cluster analyses search for groups of users according to a sets of defined dimensions. Our set of dimensions was informed by existing scales in earlier studies and adapted to suit the context of this study. In the following, we describe each of the selected metrics.

Two metrics were concerned with the characterization of preferences for places: the popularity of places in which users typically created content, and the average evaluation of places annotated by users. The popularity of a place is a function of the activity in the place, according to the GSN over a time interval. In Foursquare, the popularity of a place is reckoned according to the number of tips made by creators in that place during a period of observation. We chose tips to characterize popularity because GSNs show previous tips to the creator before they share their own tips. The popularity of places shared by a creator is then defined as the median of popularities of all places where they produce tips.

Another metric used to characterize preferences for venues is the average evaluation of places. In Foursquare, each location presents a score from 0 to 10, calculated by the GSN and based on shared information. The score is generated according to implicit and explicit feedback of content creators collaborating in venues. Our average evaluation of places where a content creator shared tips is thus the arithmetic mean score of all collaborated places by users.

Next, there were three metrics characterizing activity in space and time. The first metric, activity, was calculated as the number of tips produced by a content creator over a time interval. The second metric, variety of collaborations of content creators, gauged the number of distinct places where a content creator collaborated in Foursquare in a time interval. The third metric, distance among collaborations of the same creator, was calculated as the median of the distances, in kilometers, between each place annotated by a creator and the centroid of all venues where he or she contributed tips.

Regarding the perspective of time, we considered the periodicity of content creator activity as the median of days between consecutive collaborations of a content creator. The last metric is the social influence of a content creator. In Foursquare, a tip has three tags that work as proxies for such influences: a) likes — a tag for giving positive feedback; b) to dos — a tag for putting tips of collaborated places on a wish list; and c) lists — a tag for putting a place on a personal list. Although these tags have different purposes, they are equally important to measure the visibility and the acceptation of creators’ collaborations to other GSN users. Thus, the social influence of a content creator is defined as the mean of the ratio between the sum of likes, to dos and lists by the number of followers, as shown in Equation 1. The number of followers represents the sum of users who see a tip made by their creator, and n is the number of tips created by their creator.

equation 1


The creators in the dataset shared 6.6 tips in 1.7 places, on average. Moreover, each creator produced consecutive tips with an average time interval of 45.2 days and within an average distance of 1.5 km. On average, they preferred places with 53.9 previous tips, and with an evaluation average of 5.1 of score. The distribution of these metrics is shown in Figure 2.


Histograms of metrics about GSN content creators
Figure 2: Histograms of metrics about GSN content creators. All metrics were normalized by z-score technique in order to achieve comparable sizes.


The next step was a correlation analysis to explore pairwise variation among these metrics. The Kendall correlation coefficient was chosen in that it provided a non-parametric measure of the strength of dependence between variables, measuring dependences between the ranks. According to results as noted in Figure 3, most of the selected metrics were not correlated, except the pairs: a) periodicity and variety (correlation coefficient = -0.3), and b) distance to the centroid and activity (correlation coefficient = 0.3).


Results of Kendall correlation per pair of metrics
Figure 3: Results of Kendall correlation per pair of metrics.


The former relation illustrated, in general, that creators in Brazilian cities who visited fewer different places tended to wait for longer time intervals before sharing consecutive tips. The second correlation indicated that creators who shared more tips tended to produce tips over a wider portion of the city.

In partitional clustering, the expert must decide the number of groups. There are several heuristics for deciding what number of groups is most appropriate for analysis. This study adopted the elbow technique to define the number of groups. This method uses a visualization of the sum of squared distances between points in cluster and their centroids as the number of clusters in the analysis progresses, as shown in Figure 4A. The most appropriate number of groups was reached when the drop in the sum of distances started decreasing, forming elbows in the line such as the two highlighted in the plot. Based on results of Figure 4A, the best choice of groups was four clusters (red circle of Figure 4A). The graph also showed a second option of 13 clusters (yellow circle of Figure 4A), but we discarded this last one as it would lead to too specific clusters.


Number of clusters per average cluster distances (A) and silhouette analysis for the clustering with k=4 clusters (B)
Figure 4: Number of clusters per average cluster distances (A) and silhouette analysis for the clustering with k=4 clusters (B).


Figure 4B validated the choice of the number of clusters through a silhouette analysis. This method evaluates the similarity of one element to a cluster compared to other resulting clusters: silhouette coefficients for each point measure how far a sample is from the closest cluster to which it does not belong. A -1 coefficient states that a point is much closer to points of a cluster it does not belong to. Analogously, a coefficient of 1 represents that the point is much closer to its cluster than to other clusters. The silhouette of clusters in the 4-cluster solution suggests this number of groups provides a clear separation of content creators in four distinct profiles.




Following the suggestion of previous methods of four groups, we labeled each cluster based on the most salient characteristics of the cluster centroid. These labels are shown in Figure 5: >low frequency content creators, average content creators, mainstream content creators>, and >content producers>. To allow the comparison of metrics in different scales, the figure shows z-scores of each metric. As such, zero implies the average for that metric, and the unit is the standard deviation of that metric.


Collaborative metrics per group of GSN content creators
Figure 5: Collaborative metrics per group of GSN content creators. All metrics were normalized by z- score technique to have comparable sizes.


In the following, we detail each creator profile, according to their main features.

Low frequency content creators

The clustering process found 1,185 low frequency content creators in 21 Brazilian cities. Collectively, these creators authored 7,306 tips. When considered individually and compared to creators of other groups, the low frequency creators produced 3.43 tips, which represented the smallest number of produced tips per creator. They also spent the longest time between consecutive tips (mean = 311.5 days, sd = 142.23). Being less active, low frequency creators tended to visit a smaller variety of places (mean of 1.26 different places, sd = 0.43).

Average content creators

This group has 10,173 content creators and contained the highest number of creators in all the selected cities. Average creators were the main suppliers of geolocated content across all groups with a contribution summing 132,313 tips. According to the centroid of this group, average creators shared a total of tips near to the overall average (mean of 5.66 tips per creator, sd = 3.51) and they almost stayed in the same place to contribute (mean of 1.45 different places, sd = 0.48).

The average distance of places contributed to the centroid of collaborations was near the overall average. In other words, places with tips from a same creator had 1.53 kilometers of distance from each other, on average. Besides, these creators shared consecutive tips in a time interval less than a month (mean = 23.80 days, sd = 37.50), a little below average. Regarding spatial preferences, average creators selected places with an average popularity of 42 tips per place (sd = 57.2) and an average evaluation score of 5.1 (sd = 3.39), slightly lower than average. It meant that average creators focused on places with good evaluations to produce geolocated information. Finally, they presented a social influence lower than average (mean = 1.02, sd = 2.28).

Mainstream content creators

Mainstream creators had a marked preference for contributing in more popular places. This group had 1,829 creators in 15 Brazilian cities, producing 23,647 tips in well-visited venues. Together with their preference for popular places, these creators tended to choose places with an evaluation score lower than average (mean score = 4.89, sd = 3.40). Further, they did not follow GSN evaluations in selecting places for sharing tips. Perhaps, mainstream creators stopped at popular places because they imitated their friends or they had little knowledge of visited areas. The search for well-visited places made mainstream creators collaborate in the highest quantity of distinct places of all groups (mean of 3.35 different places, sd = 0.98).

As consequence of their preference for populated places, mainstream creators had the lowest social influence of GSN users (mean = 0.77, sd = 1.83). The creation of useful tips in popular places was a difficult process. They also produced a sum of tips below the average (mean of 5.42 tips per creator, sd = 3.4) and shared consecutive tips in a mean time interval lower than average (mean = 3.87 days, sd = 19.20). It other words, mainstream creators produced fewer tips in a shorter time. Therefore, the collaboration of mainstream creators concentrated in the smallest areas because they roamed the lowest distances of all groups (mean = 0.94 km., sd = 4.35).

Content producers

The 596 content producers of samples shared 30,676 tips in 16 Brazilian cities, excluding the cities of Brasília, Campinas, Salvador, São Bernardo do Campo and Teresina. They were the most productive collaborators (mean of 32.09 tips per creator, sd = 13.92). Also, producers collaborated in the highest quantity of different places (mean of 1.85 different places, sd = 0.64) and they spent the least time to produce consecutive tips (mean of 3.71 days, sd = 9.83).

All participation was concentrated in small areas because the average distance among visited places was lower than average (mean = 3.10 km., sd = 14.14). Moreover, producers favor places with popularity (mean = 51.7 tips, sd = 82.83) and average evaluation score (mean = 4.85 of score, sd = 3.47) to sharing tips. Thus, they represented the highest social influence (mean = 1.08, sd = 2.64) and this good acceptance of producers’ tips was an incentive for high collaboration.

Insights about groups of creators

Regarding GSNs, other features gave feedback on the visibility of contributions, namely followings and followers. The tag of followings represented the sum of tips’ authors that a creator sees in GSNs and followers are the number of users who see the collaborations of this creator. We discovered that the average content creators followed the highest number of users in Foursquare (192.61 users) and producers presented the opposite conduct (163.99 users). Perhaps, content producers learned less than average creators with shared opinions about places. Lastly, low frequency creators and mainstream creators presented a similar number of followings in Foursquare (176.46 and 177.35 followings, respectively).

Based on followers, producers had the highest number of followers (308.17 users) and low frequency creators had the lowest (185.94 users). Producers did not seek the opinions of others in Foursquare, but they had numerous followers considering their tips. Thus, the know-how present in producers’ content was sought out by GSN users. Further, low frequency creators preferred using geolocated information to sharing it in GSNs. Finally, average creators had a number of followers similar to that of mainstream creators (249.19 and 236.59 followers, respectively).

By tf-IDF (term frequency-inverse document frequency), we detected some particularities among groups related to important terms of shared content. For example, tips of four profiles considered the words “place”, “attendance” and “food” to be important terms. It seemed GSN creators in Brazilian cities felt comfortable sharing geolocated information about food as well as aspects of a place, like attendance. Tips shared by mainstream creators regarded the word “people” as a frequent term, which illustrated their interest in crowded places. Obviously, all groups included women and men as creators, as shown in Figure 6. However, men represented the majority of creators in all groups. Nonetheless, the biggest difference between genders occurred among average creators.


Distribution of cities per group of creators
Figure 6: Distribution of cities per group of creators.


We detected a variety of groups among the selected Brazilian cities, as shown in the heat map (Figure 7). Regarding details, some cities in Southeast Brazil concentrated specific types of content creators. For example, relative to tips, the activity of low frequency creators, mainstream creators and producers in Brazilian cities was more concentrated than the activity of average creators. Although average creators produced the most tips in the GSN, they were well distributed throughout the country.


Intensity of participation of low frequency creators (a), average creators (b), mainstream creators (c) and producers (d) in cities of Brazil
Figure 7: Intensity of participation of low frequency creators (a), average creators (b), mainstream creators (c) and producers (d) in cities of Brazil.


We then further checked the profiles per city, as shown in Figure 8. The graph identified that average creators are the majority in all Brazilian cities. However, producers and low frequency creators were the minority in all towns. Producers were present mostly in Manaus, Recife and Rio de Janeiro; these cities are capitals in different regions of Brazil with populations greater than 1.5 million. At last, the cities of Rio de Janeiro and São Paulo have the highest incidence of mainstream creators. These two last cities are the main metropoles of Brazil with uncountable options of crowded places in GSNs. Lastly, we identified that the city of Salvador contained just one type of creator in the entire city. It meant that the city had only average creators collaborating in the GSN.


Distribution of cities per group of content creators
Figure 8: Distribution of cities per group of content creators.


Average creators represented the most balanced group due to their attitude to create and to consume geolocated information equally. Besides, their creators trusted in GSN evaluations of places to share new tips. For system administrators, the elaboration of mechanisms may encourage the collaboration of average creators, like showing the best-evaluated places in an area when necessary.

Mainstream creators are the sniffers of popular places and they are valuable for urban analysts to map areas with a high flow of GSN users in cities. Lastly, content producers provide the main source of information in GSN. This group influences other GSN users when its creators do not follow the flow of places. Further, they are critical in all cities — due to their ability to discover new places — and a role model of collaborative behavior for all creators.

All groups of creators roamed short distances during interaction with the GSN. In the case of mainstream creators, they looked for a specific type of place that was a hot spot place in a hot spot area. Besides, low frequency and average creators collaborated little, compared with other groups. Finally, producers roamed shorter distances to share more information at the same time, compared with other groups.




GSNs play an important role in obtaining a detailed and complete snapshot of cities due to their massive scale and lack of proper tools. We found an absence of studies that privilege a variety of GSN creators in Brazilian cities. The main contribution of this study was the mapping of patterns among creators of a GSN used in Brazilian cities. We determined four groups according to collaborative metrics of creators.

Our experiments demonstrated that the challenge of investigating the diversity of content creators was also related to the correct selection of collaborative metrics. The investigation of the collaborative behavior of GSN creators opens new perspectives for profiles in a different urban context. Understanding the diversity of content creators can improve the mechanisms used in the choice of places in GSNs, such as recommendations and information filtering.

Each group was important to GSNs. Firstly, producers represented a role model of participation in GSNs. However, they were weak receivers of GSN information because they consumed little information from other creators. The presence of producers was essential to generate content in GSNs and to elaborate strategies able to change GSN creators into producers. The main information receiver was the average creator, who presented a balance between consumption and creation of geolocated content.

Mainstream creators were “mappers” of popular places but in small areas. The last group was the low frequency creators and they represented a low level of collaboration in GSNs. So, geolocated systems need elaborate mechanisms that keep low frequency creators from becoming lurkers.

We also detected a lack of creators who would travel significant distances during GSN interactions. Perhaps, this was due to socioeconomic factors in Brazilian cities. An interesting line of investigation might be the comparison of cities with different socioeconomic realities. Future work should study the profiles of content creators in Brazilian cities separately, according to collaborative metrics used in this experiment. It also seems promising to investigate creators in other Brazilian cities and verify whether the presence of the four profiles of creators found in this study are present there, too. The investigation of creators helps to understand how they deal with GSNs in different urban perspectives. Moreover, it opens new possibilities for recommending techniques relevant to the city, real-time detection of creator profiles and upgrading GSNs by leveraging awareness of their urban context. End of article


About the authors

Aline Marques Morais, Universidade Federal de Campina Grande, Brazil.
E-mail: alinemorais [at] copin [dot] ufcg [dot] edu [dot] br

Nazareno Andrade Universidade Federal de Campina Grande, Brazil.
E-mail: nazareno [at] lsd [dot] ufcg [dot] edu [dot] br



A. Abraham, A.-E. Hassanien and V. Snasel (editors), 2009. Computational social network analysis: Trends, tools and research advances. London: Springer-Verlag.
doi:, accessed 10 November 2017.

T. Agryzkov, P. Martí, L. Tortosa and J. F. Vicent, 2017. “Measuring urban activities using Foursquare data and network analysis: A case study of Murcia (Spain),” International Journal of Geographical Information Science, volume 31, number 1, pp. 100–121.
doi:, accessed 10 November 2017.

A. Bawa-Cavia, 2011. “Sensing the urban: Using location-based social network data in urban analysis,” at, accessed 10 November 2017.

d. boyd and K. Crawford, 2012. “Critical questions for big data: Provocations for a cultural, technological, and scholarly phenomenon,” Information, Communication & Society, volume 15, number 5, pp. 662–679.
doi:, accessed 10 November 2017.

C. Brown, N. Lathia, C. Mascolo, A. Noulas and V. Blondel, 2014. “Group colocation behavior in technological social networks,” PloS ONE, volume 9, number 8, e105816.
doi:, accessed 10 November 2017.

J. Chang and E. Sun, 2011. “Location3: How users share and respond to location-based data on social networking sites,” Proceedings of the Fifth International AAAI Conference on Weblogs and Social Media, pp. 74–80, and at, accessed 10 November 2017.

C. Chen, D. Zhang, B. Guo, X.; Ma, G. Pan and Z. Wu, 2015. “TripPlanner: Personalized trip planning leveraging heterogeneous crowdsourced digital footprints,” IEEE Transactions on Intelligent Transportation Systems, volume 16, number 3, pp. 1,259–1,273.
doi:, accessed 10 November 2017.

M. Chen, F. Li, G. Yu and D. Yang, 2016. “Extreme learning machine based point-of-interest recommendation in location-based social networks,” Proceedings of ELM-2015, volume 2, pp. 249–261.
doi:, accessed 10 November 2017.

M. J. Chorley, R. M. Whitaker and S. M. Allen, 2015. “Personality and location-based social networks,” Computers in Human Behavior, volume 46, pp. 45–56.
doi:, accessed 10 November 2017.

J. Cranshaw, R. Schwartz, J. Hong and N. Sadeh, 2012. “The Livehoods Project: Utilizing social media to understand the dynamics of a city,” Sixth International AAAI Conference on Weblogs and Social Media, pp. 58–65, and at, accessed 10 November 2017.

H. Gao and H. Liu, 2014. “Data analysis on location-based social networks,” In: A. Chin and D. Zhang (editors). Mobile social networking: An innovative approach. New York: Springer, pp. 165–194.
doi:, accessed 10 November 2017.

H. Gao, J. Tang, X. Hu and H. Liu, 2015. “Content-aware point of interest recommendation on location-based social networks,” AAAI’15: Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence, pp. 1,721–1,727.

J. Hardy and S. Lindtner, 2017. “Constructing a desiring user: Discourse, rurality, and design in location-based social networks,” CSCW ’17: Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing, pp. 13–25.
doi:, accessed 10 November 2017.

B. Hawelka, I. Sitko, E. Beinat, S. Sobolevsky, P. Kazakopoulos and C. Ratti, 2014. “Geo-located Twitter as proxy for global mobility patterns,” Cartography and Geographic Information Science, volume 41, number 3, pp. 260–271.
doi:, accessed 10 November 2017.

L. Hjorth, R. Wilken and K. Gu, 2012. “Ambient intimacy: A case study of the iPhone, presence, and location-based social media in Shanghai, China,” In: L. Hjorth, J. Burgess and I. Richardson (editors). Studying social media: Cultural technologies, mobile communication, and the iPhone. New York: Routledge, pp. 43–62.

C. D. F. Honig and L. MacDowall, 2017. “Spatio-temporal mapping of street art using Instagram,” First Monday, volume 22, number 3, at, accessed 10 November 2017.
doi:, accessed 10 November 2017.

L. Jin, X. Long, K. Zhang, Y.-R. Lin and J. Joshi, 2016. “Characterizing users’ check-in activities using their scores in a location-based social network,” Multimedia Systems, volume 22, number 1, pp. 87–98.
doi:, accessed 10 November 2017.

J. Kang and K. Lerman, 2011. “Leveraging user diversity to harvest knowledge on the social Web,” 2011 IEEE Third International Conference on Privacy, Security, Risk and Trust (PASSAT) and 2011 IEEE Third Inernational Conference on Social Computing (SocialCom), pp. 215–222.
doi:, accessed 10 November 2017.

Y. Li, M. Steiner, L. Wang, Z.-L. Zhang and J. Bao, 2012. “Dissecting Foursquare venue popularity via random region sampling,” CoNEXT Student ’12: Proceedings of the 2012 ACM Conference on CoNEXT Student Workshop, pp. 21–22.
doi:, accessed 10 November 2017.

Y. Liang, J. Caverlee, Z. Cheng and K. Y. Kamath, 2013. “How big is the crowd? Event and location based population modeling in social media,” HT ’13: Proceedings of the 24th ACM Conference on Hypertext and Social Media, pp. 99–108.
doi:, accessed 10 November 2017.

A. Morais, 2015. “Spatial analysis about users collaboration on geo-social networks in a Brazilian city,” International Journal on Web Service Computing, volume 6, number 4,, accessed 10 November 2017.

A. Morais and N. and Andrade, 2014. “The relevance of annotations shared by tourists and residents on a geo-social network during a large-scale touristic event: The case of São João,” In: C. Rossitto, L. Ciolfi, D. Martin and B. Conein (editors). COOP 2014 — Proceedings of the 11th International Conference on the Design of Cooperative Systems, 27-30 May 2014, Nice (France). Cham, Switzerland: Springer International, pp. 393–408.
doi:, accessed 10 November 2017.

A. Noulas, S. Scellato, C. Mascolo and M. Pontil, 2011. “Exploiting semantic annotations for clustering geographic areas and users in location-based social networks,” Social Mobile Web: Papers from the 2011 ICWSM Workshop,, accessed 10 November 2017.

D. Preoţiuc-Pietro and T. Cohn, 2013. “Mining user behaviours: A study of check-in patterns in location based social networks,” WebSci ’13: Proceedings of the 5th Annual ACM Web Science Conference, pp. 306–315.
doi:, accessed 10 November 2017.

S. M. Rahimi and X. Wang, 2013. “Location recommendation based on periodicity of human activities and location categories,” In: J. Pei, V. S. Tseng, L. Cao, H. Motoda and G. Xu (editors). Advances in Knowledge Discovery and Data Mining. Lecture Notes in Computer Science, volume 7819. Berlin: Springer, pp. 377–389.
doi:, accessed 10 November 2017.

D. Sarkar, R. Sieber and R. Sengupta, 2016. “GIScience considerations in spatial social networks,” In: J. A. Miller, D. O’Sullivan and N. Wiegand (editors). Geographic information science. Lecture Notes in Computer Science, volume 9927. Cham, Switzerland: Springer International, pp. 85–98.
doi:, accessed 10 November 2017.

A. K. Schaar, A. C. Valdez and M. Ziefle, 2013. “The impact of user diversity on the willingness to disclose personal information,” In: A. Holzinger, M. Ziefle, M. Hitz and M. Debevc (editors). Human factors in computing and informatics. Lecture Notes in Computer Science, volume 7946. Berlin: Springer, pp. 174–193.
doi:, accessed 10 November 2017.

T. H. Silva, P. O. Vaz de Melo, J. M. Almeida, J. Salles, A. A. F. Loureiro, 2013a. “Social media as a source of sensing to study city dynamics and urban social behavior: Approaches, models, and opportunities,” In: M. Atzmueller, A. Chin, D. Helic and A. Hotho (editors). Ubiquitous social media analysis. Lecture Notes in Computer Science, volume 8329. Berlin: Springer, pp. 63–87.
doi:, accessed 10 November 2017.

T. H. Silva, P. O. Vaz de Melo, J. M. Almeida, J. Salles, A. A. F. Loureiro, 2013b. “A picture of Instagram is worth more than a thousand words: Workload characterization and application,” DCOSS ’13: Proceedings of the 2013 IEEE International Conference on Distributed Computing in Sensor Systems, pp. 123–132.
doi:, accessed 10 November 2017.

T. H. Silva, P. O. Vaz de Melo, J. M. Almeida and A. A. F. Loureiro, 2014. “Large-scale study of city dynamics and urban social behavior using participatory sensing,” IEEE Wireless Communications, volume 21, number 1, pp. 42–51.
doi:, accessed 10 November 2017.

Statista, 2014. “Number of smartphone users worldwide from 2014 to 2020 (in billions),” at, accessed 18 February 2017.

S. M. Straathof, 2007. “Shannon’s entropy as an index of product variety,” Economics Letters, volume 94, number 2, pp. 297–303.
doi:, accessed 10 November 2017.

H. Weber and J. Novet, 2015. “Foursquare by the numbers: 60M registered users, 50 MAUs, and 75M tips to date” (18 August), at, accessed 10 November 2017.

W. Wörndl, A. Hefele and D. Herzog, 2017. “Recommending a sequence of interesting places for tourist trips,” Information Technology & Tourism, volume 17, number 1, pp. 31–54.
doi:, accessed 10 November 2017.

D. Wu, N. Mamoulis and J. Shi, 2015. “Clustering in geo-social networks,” Bulletin of the IEEE Computer Society Technical Committee on Data Engineering, at, accessed 10 November 2017.

F. Wu and Z. and Li, 2016. “Where did you go: Personalized annotation of mobility records,” CIKM ’16: Proceedings of the 25th ACM International on Conference on Information and Knowledge Management, pp. 589–598.
doi:, accessed 10 November 2017.

J. Yang, C. Hauff, G.-J. Houben and C. T. Bolivar, 2016. “Diversity in urban social media analytics,” In: A. Bozzon, P. Cudre-Maroux and C. Pautasso (editors). Web engineering. Lecture Notes in Computer Science, volume 9671. Chaum, Switzerland: Springer International, pp. 335–353.
doi:, accessed 10 November 2017.

C. Yu, B. Xiao, D. Yao, X. Ding and H. Jin, 2017. “Using check-in features to partition locations for individual users in location based social network,” Information Fusion, volume 37, number C, pp. 86–97.
doi:, accessed 10 November 2017.

J. Zhang, X. Kong and P. S. Yu, 2016. “Social badge system analysis,” 2016 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), pp. 453–460.
doi:, accessed 10 November 2017.


Editorial history

Received 4 June 2017; revised 30 August 2017; accepted 10 November 2017.

Creative Commons License
“Profiles of mapping of content creators in a geo-social network: The case of 21 Brazilian cities” by Aline Morais and Nazareno Andrade is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Profiles of mapping of content creators in a geo-social network: The case of 21 Brazilian cities
by Aline Marques Morais and Nazareno Andrade.
First Monday, Volume 22, Number 12 - 4 December 2017

A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2019. ISSN 1396-0466.