SPATIAL ANALYSIS ABOUT USERS COLLABORATION ON GEO-SOCIAL NETWORKS IN A BRAZILIAN CITY

International Journal on Web Service Computing (IJWSC), Vol.6, No.4, December 2015
DOI : 10.5121/ijwsc.2015.6401 1
SPATIAL ANALYSIS ABOUT USERS COLLABORATION
ON GEO-SOCIAL NETWORKS IN A BRAZILIAN CITY
Aline Morais
1
Universidade Federal de Campina Grande, Brasil
ABSTRACT
Geo-Social Networks (GSNs) are collaborative systems that has the geolocated information as main
component. The geolocation resource integrates virtual and real worlds, allowing the comprehension about
these two scenarios at same time. Based on that, this work define a process of spatial analysis of shared
information on a GSN. The present work proposes the usage of six spatial features as feedback about
collaborative behaviour on city. The spatial analysis aims understand if users’ collaboration change
among city census sectors. Understanding how users deal with GSNs in an area, will help about
collaborative patterns per urban region. As result, this work detected spatial patterns among users in the
GSN Foursquare of a Brazilian city. These patterns indicates that users’ collaboration receive influences of
extrinsic and intrinsic features of GSN and the comprehension about their users is a complex task.
KEYWORDS
Geo-Social Networks, Spatial Analysis, Collaborative Behaviour
1. INTRODUCTION
The Geo-Social Networks (GSNs) are social networks that share geolocated information, based
on location of visited places by their users. The analysis of interactions in GSN can show how
users collaborate in a place. Many investigations about GSNs has explored spatial analysis in
order to map preferences and patterns of displacement of users. Other type of investigation about
GSNs is about spatial analysis focused on a delimitated area, as a city [2]. This last line of
investigation is few explored because of complexity to mining geolocated data in a specific
region.
Based on that, the present work has as research problem the absence of investigation about
collaborative behaviour of GSN users in Brazilian cities. The acceptation of GSNs in Brazil
increase considerably in last years. At last, this work has as main aim the definition of spatial
metrics able to measure the collaborative behaviours in GSNs. The data used to calculate these
metrics is tips shared on GSN Foursquare. It is possible because tips have geographic coordinates
integrated with their content.
The metrics observed in the present work are: a) coverage, b) proximity, c) spreading, d) activity,
e) diversity and f) actuality. The coverage measures the percentage distribution in a region. The
proximity evaluates the distance among tips shared consecutively on time and in a same region.
Spreading represents the distance among tips in a same area, without temporal reference. Activity
is the number tips in an area. Diversity corresponds to a number of different tips' authors in a
same area. Lastly, the actuality represents the lifetime of tips in a region. At last, this work
observed these metrics in subareas of city, called to census sectors.

2
This paper proposes methods to analyze spatially the GSN interaction over a city. First, the
Section 2 describes the related works about spatial analysis in social networks. At next, the
present work will explain about methodology adopted to describe these metrics over a Brazilian
city. This section will also describe the first results about the metrics. At last, the section 4 will
summarize about main findings of this work.
2. RELATED WORKS
GSN are social networks that qualify location as the core object of its structure [3]. There are
investigations with GSNs that adopt the association between urban features and GSN features in
order to understand human relationships and interactions [4] [5]. Related to the association
between urban features and GSN features, it has a few works that observed it [6] [7]. Many of
these works aim to extract Points of Interested (POIs) of its users in specific areas, mainly for
recommendation techniques. However, the present paper found few works that focus on gap
between digital worlds and socioeconomic factors about localities [19].
There are many metrics on GSN able to have relation with socioeconomic features on a city.
Related to that, the present work shall highlight features used by it. The first metric is the
diversity and it represents how users may vary based on collaboration [9] and origin [10], for
example. Other metric used broadly is the activity. Normally this metric measures the users’
participation on collaborative systems [8]. Related to coverage metric, some investigations adopt
this metric in order to understand better about the percentage of some variables in a context
generic. In collaborative systems, the coverage is useful to measure percentage of participation to
associate with other occurrences outside of systems [11].
Some previous works highlight the importance to consider metrics with distance during analysis
of features on collaborative systems [12]. Because of that, the present work will use two distance
metrics on GSN: spreading and proximity. Spreading is a metric that measures the space among
points in an area [13]. The proximity measures data as the number or frequency of contacts
between individuals, the size of share-holding relations between enterprises, the number of
members in common between organizations, and so on [14]. The last metric is the actuality and it
represents the novelty degree shared in collaborative systems. Normally, this metric is based on
lifetime features [15].
3. METHODS AND RESULTS
This work developed a longitudinal study, which aimed to identify peculiarities about the
collaborative participation of GSN users over regions of a Brazilian city. To understand it, the
present paper chose Campina Grande city to be part of analysis. It is a medium Brazilian city with
around 626,937 people and it is in Northeast of Brazil [20]. Moreover, this work observed areas
of Campina Grande, called census sectors. They are political-administrative divisions that
represent territorial units, created and registered by Instituto Brasileiro de Geografia e Estatística
(IBGE – Statistics and Geography Brazilian Institute). The Campina Grande city has 439 census
sectors, only in the urban area. At last, the chosen GSN for this work is the Foursquare. It is one
of the most popular GSNs on Brazil in 2014. Furthermore, this system present 40 million users
over the world.
As mentioned previously, the spatial metrics are focused on tips shared by GSN Foursquare. Tips
represent annotations created by users in order to comment publicly about visited places in GSN.
The present work analysed tips shared in census sectors of Campina Grande during March, April
and May of 2014. The present work define the spatial metrics based on tips features, at following:

3
• Diversity (user per sector): mean value about number of users in a census sector;
• Activity (tips per sector): number of tips in a census sector;
• Actuality (days): mean of lifetime of tips in a census sector;
• Coverage (%): distribution percentage about tips in a census sector;
• Spreading (meters): distance among users in a census sector;
• Proximity (m): spatial distance among tips of different places in a census sector.
This work observes spatial patterns with assistance of ArcGIS 10.1 software, mainly to plot GSN
metrics in census sectors and to calculate the metrics. With this software, the present work
executes three geographic operations able to distinguish the presence of spatial patterns. The
spatial techniques used in investigation are:
• Spatial autocorrelation (Moran’s I): It measures and analyzes the degree of
dependency among tips in a census sector. This method shows measures how much
close objects are in comparison with other close objects [16].
• Hot Spot Analysis (Getis Ord Gi*): Hot Spot Analysis is a detection method of
spatial clusters, which identifies spatial concentrations, statistically significant, of the
high and of low values associated with a set of geographic features[17];
• Cluster and Outlier Analysis (Anselin Local Moran’s I): Given a set of weighted
features, identifies where high or low values cluster spatially, and features with
values that are very different from surrounding feature values [18].
3.1. Diversity
The diversity of tips reflects the popularity of places in an urban region. This work has as
suspicion that sectors with high number of users per place, also present: a) Popular places; b)
Places with updated tips; c) Places with tips’ collaborators close among them. The Figure 1(a)
details about diversity is in Campina Grande. The darker the sector, the greater the diversity. The
darkest areas are located mainly in center of Campina Grande. It is possible realize the variation
about diversity indexes over census sectors of the Brazilian city.
Firstly, it investigates if census sectors with same diversity index are near spatially on the city. To
solve that the present work executed spatial auto-correlation test of Moran’s I, as explained
previously. If autocorrelation exists in a map, then this violates the fact that observations are
independent from one another. The Moran’s I index is a value between -1 (disperse) and 1
(grouped). According with the result in Figure 1(b), it is not possible say the number of different
users per census sector (diversity) presents a dispersed or grouped behaviour over Campina
Grande city, due to p-value higher than 0.05.

4
Figure 1. (a) Diversity in Campina Grande. (b) Result of spatial autocorrelation about Diversity.
The next step was the local analysis about diversity on census sectors, after detect the inexistence
of a global pattern among census sector. Thus, this investigation applied Anselin Moran’s I
method as an indicator of spatial association. The results is in Figure 2 (a).
By the map, the majority of census sectors present not significant index compared with other
diversity values in census sectors. The next procedure is the usage of a method able to identify
groups, according to diversity on census sectors. The chosen method was hotspot analysis based
on Getis-Ord Gi index. The hotspot analysis calculates the statistics about Gets-Ord Gi index and
GiZ-score informs about groups with high and low values spatially. Observing the map, the
central area presents a group of census sectors with the biggest quantity of different users in the
same area (red sectors in Figure 2b). In minority, it saw a few sectors with diversity lower than
average (blue sectors in Figure 2b). The yellow sectors represent diversity values without
statistical significance.
Figure 2. (a) Result of group mapping with Anselin Moran’s I about Diversity. (b) Result of Hot Spot
analysis about diversity.

5
If a census sector has a high diversity and does not present popular places, probably the census
sector present new places to visit. Related to this, it is an urban area with many novelties shared
on GSN. Thus, novelty gives seasonality to services and features of places. At last, this work will
investigate this possibility through actuality investigation, in next sections.
3.2. Activity
The activity represents the number of tips shared per square kilometer in each census sector of
Campina Grande city. The initial suspicious is census sectors, which present places with high
activity, also have places with recent tips on it. The description about activity on census sectors is
in Figure 3 (a).
The activity is higher in areas located on the city center. However, the major part of census
sectors did not present any activity, as it shows the clearest areas of Figure 3(a). As in diversity
perspective, this paper investigated about spatial patterns related to activity index over census
sectors. Again, it did a spatial autocorrelation of Moran’s I, in Figure 3(b).
Figure 3. (a) Activity over census sectors of Campina Grande. (b) Result of spatial autocorrelation about
Activity.
The Moran’s index (I = 0.10) shows a spatial pattern slightly aggregated based on activity index
over census sectors. It presents significant results due to the p-value less than 0.05 and the z-score
higher than 0.96. These results mean that the activity about creation of geolocated content over
Campina Grande are dependent to other factors.
After detect the presence of global spatial patterns about the number of tips per square kilometer,
it is necessary evaluate on a local scale. Then, it applied Anselin Moran’s I (Figure 4a) and Hot
Spot analysis (Figure 4b).
According to Anselin Moran’s I results, the activity behaves differently over census sectors. For
example, there are some areas with high contribution of tips per squared kilometer, rounded by
sectors with high activity (black sectors in Figure 4a). According to Figure 4(b), there are two
main groups based on activity: sectors with activity higher than average (sectors in red colour)
and sectors with activity lower than average (sectors in blue colour). In this last group, there is
only one sector.

6
Figure 4. (a) Result of group mapping with Anselin Moran’s I about Activity. (b) Result of Hot Spot
analysis about activity.
Therefore, some sectors with high activity in Hot Spot Analysis are not significant in Anselin
Moran’s I. It happens because these sectors are spatially isolated. In this sense, the center of
Campina Grande has the highest quantity of tips per square kilometer with some areas of high
activity isolated among other census sector with high activity.
3.3. Coverage
The coverage has as aim to measure the distribution of tips on census sector. The suspicious is
sectors with high coverage, also present recent tips. According to Figure 5 (a) the coverage in
Campina Grande is heterogeneous over census sectors. As happened with two previous metrics,
there is a presence of high indexes in the center of Campina Grande city. Moreover, the majority
of sectors present coverage near to 0% between March and May of 2014.
Figure 5. (a) Coverage over census sectors of Campina Grande. (b) Result of spatial autocorrelation about
Coverage.

7
Again, this work applied the spatial autocorrelation in order to understand the existence or not of
patterns on Campina Grande city. According to results of Figure 5 (b), it is present a slightly
clustered behaviour related to coverage metric over census sectors. Related to initial results about
presence of patterns, the present investigation applied in sequence the Anselin Moran’s I (Figure
6a) and Getis-Ord Gi* (Figure 6b) methods in order to obtain details about these patterns.
Figure 6. (a) Result of group mapping with Anselin Moran’s I about Coverage. (b) Result of Hot Spot
analysis about Coverage.
The results about Anselin Moran’s I method presented, until now, the best quantity of sectors
with high coverage (HH Legend in Figure 6a). All of these sectors is concentrated in the center of
Campina Grande city. One part of Liberdade district has low coverage (LL Legend on Figure 6a).
Moreover, part of Palmeira, Conceição and São José district present a low coverage rounded by
sectors with high coverage (LH Legend on Figure 6a).
According to results of Getis-Ord Gi* method, it is possible to distinguish two main groups. The
former belongs to sectors with a coverage higher than average (red sectors in Figure 6b). The
latter group is about sectors with coverage lower than average (blue sectors in Figure 6b). The
blue sectors are part of districts belonging to Liberdade, Louzeiro, Jardim Continental, among
others.
3.4. Actuality
The actuality measures how new are the shared tips per sector. The first impression is census
sectors with high actuality also have an area with popular places, collaborative users and a high
coverage. The panorama about actuality metric in Campina Grande city is showed in Figure 7(a).

8
Figure 7. (a) Actuality over census sectors of Campina Grande. (b) Result of spatial autocorrelation about
Actuality.
According to Figure 7(a), it is possible realize a centralized behaviour at the first moment. In this
case, clear sectors mean sectors with the best index of actuality. In this work, the actuality is a
measure between 69 and 159 days of the lifetime. It happens because has a relation with
collecting date and mining date.
Even as previous features, the present work did a spatial autocorrelation to get more results about
spatial patterns. According to results of Figure 7b, it cannot say anything about actuality in
Campina Grande sectors because the result do not have statistical significance (p-value > 0.05).
However, the local spatial analysis is able to understand the actuality over census sectors. Based
on that, it executed Anselin Moran’s I method. As explained previously, this method helps to
identify outliers and spatial groups, according to a specific feature. The results are in Figure 8a.
Figure 8. (a) Result of group mapping with Anselin Moran’s I about Actuality. (b) Result of Hot Spot
analysis about Actuality.

9
The results of Figure 8(a) showed that there is not big cluster related to actuality in Campina
Grande city. It has three sectors (referee to Catolé and Estação Velha districts) that present high
values of actuality (HH Legend in Figure 8a). In other words, these sectors present places with a
little novelty. Perhaps, the tips on these places did not update. The sector belongs to Centenário
district present a high value related to lifetime, but it is rounded by others sectors with younger
tips (HL Legend in Figure 8a). At last, the Santa Rosa and Centenário districts presented the
youngest tips compared to all sectors (LL Legend in Figure 8a).
It found out that census sectors of a same district present opposite behaviour related to actuality
metric. It happens in Centenário district. The north part of this district presented ancient tips and
the south area presented younger tips. It happens because the north area is a residential area while
the south area has banks and restaurants that attract people flow. The next step is mapping
significative groups based on tips actuality. To do this, it ran Getis-Ord Gi* method (Figure 8b).
According to results of Figure 8(b), it detects the absence of concentrated areas related to
actuality metric in Campina Grande city. It detected some sectors belonging to Catolé, José
Pinheiro and Sandra Cavalcante districts, which presented values higher than average of tips.
Otherwise, the part of Santa Rosa district presented sectors with tips more recent.
3.5. Spreading
The spreading reflects how consecutives tips are near overtime. Census sectors with low
spreading probably will present popular places, places with high participation, places with new
tips and low coverage. The general configuration about spreading is showed in Figure 9(a).
Figure 9. (a) Spreading over census sectors of Campina Grande. (b) Result of spatial autocorrelation about
Spreading.
The sectors with the highest spreading are belonging to districts of Bela Vista (3.3 meters), Catolé
(8.2 meters), Centenário (19.4 meters) and Center (19.6 meters). Otherwise, there are some
sectors with spreading near to 0 meters because they have tips created in the same or near places.
It happens in districts of Liberdade, São José, Palmeira, and Conceição, for example. The sectors
with tips shared more distant among themselves are at Bairro Universitário, Centro and
Dinamérica districts.

10
According to results in Figure 9(b), it is not possible conclude anything about Moran’s I spatial
autocorrelation because the results are not statistical significance (p-value > 0.05 and z-score <
1.96). Although it did not detect a global pattern, it is possible mapping cluster in a local scale by
Anselin Moran’s I method, as seen in Figure 10 (a).
Figure 10. (a) Result of group mapping with Anselin Moran’s I about Spreading. (b) Result of Hot Spot
analysis about Spreading.
Based on results of Figure 10 (a), it is possible to observe two main groups. The former is sectors
with a long distance among tips (legend black in Figure 10 (a). The latter is three sectors with a
long distance among their tips and rounded by sectors with opposite behaviour. The sectors
correspond to part of Bairro Universitário, Dinamérica and Sandra Cavalcanti districts.
At next, it will see this pattern based on Getis-Ord Gi* method. Based the results of Figure 10 (b),
it detected the presence of clusters on central area of city and in Bairro Universitário district.
They present high level of spreading in sectors. At last, a part of Jardim Paulistano district has
tips near among their tips
3.6. Proximity
The proximity is a spatial metric that aims understand distance among tips in different places. Its
formula is a ratio between the mean of distance and quantity of places in each sector. The
panorama is in Figure 11 (a).
Based on results of Figure 11 (a), it can see a high distance among places mainly in the district of
Bairro Universitário. It happens because this district presents few constructions, especially two
universities. The next stage is the execution of spatial autocorrelation in order to evaluate the
patterns about proximity metric (Figure 11b). The result of p-value about Moran’s I method did
not reveal statistical significance on results.

11
Figure 11. (a) Proximity over census sectors of Campina Grande. (b) Result of spatial autocorrelation about
Proximity.
The other way to find patterns about proximity is applying Anselin Moran’s I method. Based on
the results of Figure 12(a), it highlights only four sectors. A part of the district of Prata and
Centro present high distance among tips in different places. At last, a part of Quarenta and
Dinamérica districts have a high value of proximity, rounded by sectors with small values of
proximity among places.
Figure 12. (a) Result of group mapping with Anselin Moran’s I about Proximity. (b) Result of Hot Spot
analysis about Proximity.
At next, the present paper executed Getis-Ord Gi* method to mapping better these sectors. Based
on results of Figure 12(b), it detected a cluster with proximity value lower than the average of
sectors. This cluster is represented by parts of Prata, Centro e São José districts.

12
4. CONCLUSIONS
Related to diversity, the maps showed that sectors, which present high quantity of users per
sector, also have users so near among them, a high coverage of tips and a high number of tips per
square kilometer. Census sectors with high activity also are sectors with a high number of tips per
square kilometer and a high number of tips per place. Moreover, sectors with high activity also
present high coverage and distance among consecutive tips.
Related to coverage, this work found out sectors of Campina Grande, which present high
coverage also present a high activity, generally concentrated in the city center. Based on the age
of tips per sector (actuality), the present paper found out that census sectors with novel tips are
areas with a low number of tips per place (activity).
Comparing maps of spreading and proximity, it is possible realize some sectors with low
spreading because their tips belonged to the same place. For example, some sectors in the city
center were much tagged in popular places. Otherwise, a part of Dinamérica district showed tips
shared in different places but near among them. Then, it shows how important consider spreading
and proximity during the spatial analysis process.
Another interesting fact is during comparisons between spreading and proximity in an area of
Campina Grande, called Açude Velho. The spreading index in this area is high and it happens due
to two possible reasons. The first is because consecutive tips are done in opposite sides of the
area. The second possible reason is due to physical impediment. Observing proximity index, it
confirms the second possibility as true. It happens because this work realized that this spreading
index is homogeneous among census sectors on that region.
Finally, sectors with high proximity, also present a high number of tips per square kilometer and a
high coverage. For next phase of this research, this work will compare these results with other
cities.
Thus, the results about the present investigation reinforce the importance for understanding urban
context. As the future works, it will map which urban factors are able to affect these different
behaviors about GSN collaboration over the city. The second planning is understand how this
comprehension may be useful for other GSN users, during the decision making process. Probably,
it will be through recommendation techniques based on the kind of place that a GSN user want
visit.
ACKNOWLEDGEMENTS
Aline Morais thanks CAPES for financial support.
REFERENCES
[1] Bao, J., Zheng, Y., Wilkie, D., & Mokbel, M. (2015). Recommendations in location-based social
networks: a survey. GeoInformatica, 19(3), 525-565.
[2] Jiang, B., & Miao, Y. (2015). The evolution of natural cities from the perspective of location-based
social media. The Professional Geographer, 67(2), 295-306.
[3] Symeonidis, Panagiotis, Dimitrios Ntempos, and Yannis Manolopoulos. Recommender Systems for
Location-based Social Networks. Springer, 2014.
[4] Davis Jr, C. A., Pappa, G. L., de Oliveira, D. R. R., & de L Arcanjo, F. (2011). Inferring the location
of twitter messages based on user relationships. Transactions in GIS, 15(6), 735-751.

13
[5] Silva, A. D. S. E. (2008). Alien revolt (2005-2007): A case study of the first location-based mobile
game in Brazil. Technology and Society Magazine, IEEE, 27(1), 18-28.
[6] Pinquart, M., & Sörensen, S. (2000). Influences of socioeconomic status, social network, and
competence on subjective well-being in later life: a meta-analysis. Psychology and aging, 15(2), 187.
[7] Gao, H., Tang, J., Hu, X., & Liu, H. (2015, February). Content-aware point of interest
recommendation on location-based social networks. In Proceedings of the 29th AAAI Conference on
Artificial Intelligence.
[8] Viren, P. P., Vogt, C. A., Kline, C., Rummel, A. M., & Tsao, J. (2015). Social network participation
and coverage by tourism industry sector. Journal of Destination Marketing & Management.
[9] Weng, L., & Menczer, F. (2015). Topicality and Impact in Social Media: Diverse Messages, Focused
Messengers. PloS one, 10(2), e0118410.
[10] Morais, A., & Andrade, N. (2014, January). 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
COOP 2014-Proceedings of the 11th International Conference on the Design of Cooperative Systems,
27-30 May 2014, Nice (France) (pp. 393-408). Springer International Publishing.
[11] Ellison, N. B. (2007). Social network sites: Definition, history, and scholarship. Journal of
Computer Mediated Communication, 13(1), 210-230.
[12] Bjørn, P., Esbensen, M., Jensen, R. E., & Matthiesen, S. (2014). Does distance still matter? Revisiting
the CSCW fundamentals on distributed collaboration. ACM Transactions on Computer-Human
Interaction, 21(5), 1-27.
[13] Wong, L. H., Pattison, P., & Robins, G. (2006). A spatial model for social networks. Physica A:
Statistical Mechanics and its Applications, 360(1), 99-120.
[14] Scott, J., & Carrington, P. J. (2011). The SAGE handbook of social network analysis. SAGE
publications.
[15] Wilson, C., Boe, B., Sala, A., Puttaswamy, K. P., & Zhao, B. Y. (2009, April). User interactions in
social networks and their implications. In Proceedings of the 4th ACM European conference on
Computer systems (pp. 205-218). Acm.
[16] Carrington, Peter J., John Scott, and Stanley Wasserman, eds. Models and methods in social network
analysis. Vol. 28. Cambridge university press, 2005.
[17] Grubesic, T. H. (2006). On the application of fuzzy clustering for crime hot spot detection. Journal of
Quantitative Criminology, 22(1), 77-105.
[18] Shaker, R., Craciun, A., & Gradinaru, I. (2010). Relating land cover and urban patterns to aquatic
ecological integrity: a spatial analysis. Geographica Technica, 1, 76-90.
[19] Sales, A., Alves, L., Araújo, M., Menezes, A., Morais, A., & Andrade, N. (2013, October). O uso de
uma rede geossocial nas cidades brasileiras e sua relação com fatores socioeconômicos. In
Proceedings of the 12th Brazilian Symposium on Human Factors in Computing Systems (pp. 142-
147). Brazilian Computer Society.
[20] IBGE- Instituto Brasileiro de Geografia e Estatística. Available at:
http://cidades.ibge.gov.br/painel/painel.php?codmun=250400. Acessed in June 2015.
AUTHOR
Aline Morais
She is PhD student at Universidade Federal de Campina Grande (UFCG). She was a
professor at Instituto Federal de Educação, Ciência e Tecnologia da Paraíba (IFPB) in
courses of Electrical Engineering, GIS Technology, Computer Networking Technology,
Systems in Telecommunications Technology and Systems for the Internet Technology. She
is technologist in Computer Networking, by IFPB, and is Msc. in Informatic at Universidade
Federal da Paraíba (UFPB). Nowadays, she is in Laboratório de Sistemas Distribuídos
(LSD), in UFCG. Recently, she has collaborated with L3S Reseach Center (Hannover,
Germany) for working in PhD sandwich.

SPATIAL ANALYSIS ABOUT USERS COLLABORATION ON GEO-SOCIAL NETWORKS IN A BRAZILIAN CITY

More Related Content

Similar to SPATIAL ANALYSIS ABOUT USERS COLLABORATION ON GEO-SOCIAL NETWORKS IN A BRAZILIAN CITY (20)

Recently uploaded (20)

SPATIAL ANALYSIS ABOUT USERS COLLABORATION ON GEO-SOCIAL NETWORKS IN A BRAZILIAN CITY