Using Location-Based Social Media Data to Observe Check-In Behavior and Gender Difference: Bringing Weibo Data into Play

: Population density and distribution of services represents the growth and demographic shift of the cities. For urban planners, population density and check-in behavior in space and time are vital factors for planning and development of sustainable cities. Location-based social network (LBSN) data seems to be a complement to many traditional methods (i


Introduction
Personal behavior and characteristics are intimately intertwined with city planning and human mobility [1] although, in past, many traditional methods (i.e., survey, census) are used to collect data about human mobility and population density, but these traditional methods are expensive and require more processing time, produce sparse data and not that effective in policymaking.
With the introduction of LBSN's (i.e., Weibo [2], Facebook [3], Twitter [4]), users can share their location as well as the activity (referred as "check-in" [5]). Sharing check-ins allows users to announce and discuss places they visit (e.g., eating at local restaurants, shopping, visiting popular area) as part of their social interaction online. This check-in phenomenon and fast sharing of information have attracted more than 222 million subscribers. Statistics showed there were 500 million users with more than 100 million daily users on Weibo by the third quarter of 2015 [6]. These activities generate an enormous amount of users data (also referred "Big Data" [7]) based on human mobility. Despite some limitations on representing check-in behavior, e.g., bias of gender, a low sampling frequency, and bias of location category, check-in data has the ability to uncover check-in behavior within a city. Compared to the aforementioned traditional methods, LBSN data are highly available and low cost. Moreover, this data contains rich information about geolocation [8], which can be used to study check-in behavior. Thus, geo-location data offers new dimensions toward studying check-in behaviors and helps to create new techniques and approaches to analyze LBSN data. Moreover, it seems that LBSN data can be a supplement to than a substitute of traditional data sources for policy making [9]. Therefore, LBSN data can be considered as a supplement while taking policy decision related to urban planning and public services by identifying the sentiment about a topic or community detection and user analysis for identification of the actors involved [10][11][12][13][14][15][16].
In this research, we reconnoiter the reasonable prospect of using LBSN data as a novel perspective to observe individual level check-in behavior and intensity of check-ins during the period within a city. We will explore check-in behavior in 10 districts (Baoshan, Changning, Hongkou, Huangpu, Jingan, Minhang, Pudong New Area, Putuo, Xuhui, and Yangpu) of Shanghai, China, which are interconnected to the boundaries of the city center. We discuss an empirical exploration using Weibo (launched by Sina Corporation on 14 August 2009) dataset, which is a dominant social media site in China. Since each Weibo account carries information about the gender of the user, we can differentiate between LBSN usage behavior by males and females. Furthermore, we consider LBSN data can be helpful to observe check-in frequencies during weekday and weekend.
The rest of the paper is organized as follows. Section 2 overviews related works. Section 3 describes the study area and data set used in the current study. Section 4 presents the methodology. Section 5 presents the results and discussion for the experimental results performed on dataset. Finally, Section 6 concludes the paper and proposes some further research issues.

Related Work
Studying people's behavior toward services has long been constrained to analyze traditional datasets due to enhanced capabilities of capturing, analyzing, and processing geo-location data, and the field of spatial analysis has blossomed [17]. The origin of social networks lies in the early 1990s with simple communication mechanism to meet people over the internet, where people could exchange ideas. The term "social network site" (SNS) refers to web-based services. It gives people three significant capabilities: (1) to construct a public or semi-public profile, (2) to identify a list of other users with whom a connection shared, and (3) to view and track individual connections and those made by others within the system [18].
When SNSs first emerged, they were only accessible through personal computers [19]. However, recent technological advancements of "smart" mobile devices have allowed users to access their social network accounts in fixed as well as mobile stations on the move. While users have the option to access, communicate, and exchange information on SNSs via their personal computer [20], the options to access SNSs on smartphones has allowed them to easily and conveniently communicate with their "friends" at any time, anywhere [21]. As mobile development continues to progress, users share information (text, audio, video) which contain location-specific information, i.e., geo-location. With rapid use of smartphones in the recent decade, the significant innovation is the geo-location capabilities, prompting the rise and commercialization of location-based services (LBSs) [22]. Sharing information is not only just about what users are doing; it is also about what, where, why and whom they are sharing. Integration of technologies drove the development of LBSNs. LBSNs are a type of social networking in which geographic services and capabilities such as geocoding and geo-tagging are used to enable additional social dynamics [23,24]. LBSNs allow users to share their current geo-location and see their friends' location, which opens the debate about user's privacy. Privacy in LBSN is not necessarily an individual issue but extends to organizational and institutional actors involved in data sharing [25]. Some of the private data are shared by the user unsuspectingly or voluntarily. Sometimes, information is intentionally shared by the users are extracted from them extrinsically by offering them some benefits. Through the location based social network Service (LBSNS) like FireEagle, Google Latitude, Wechat, Nearby etc. are able to identify the location of a person. Some are even able to identify the location of his/her friends [26].
Various studies have been conducted to study check-in behavior under different perspectives like privacy [27,28], gender differences [29], and geographical distances [30]. Research [31,32] has found that the capacity of sharing information with millions of users is a simple method to meet with friends, make new friends, experience new things, and manage one's identity. Zheng, et al. [33] Designed an approach to mine the correlation between locations from a large amount of people's location histories. Beyond the geo-distance and the category relationship between locations, the correlation describes a more comprehensive relationship between locations in the space of human behavior and is a more nature way for human understanding. Comito, et al. [34] Presented a novel methodology to extract and analyze the time-and geo-references associated with social data so as to mine information about human dynamics and behaviors within urban context. In another study [35] presented a cloud-based software environment specifically designed for urban computing supporting smart city applications and described in detail the design and workflow for the implementation of the application and its execution by a workflow engine integrated in the environment. Brimicombe and Li [36] developed city intelligence idea that measures city ability to produce favorable conditions to get metropolitan operators (i.e., inhabitants, systems, and public/private groups) and Cheng, et al. [37] investigated the interrelation between the smart city and urban planning. Also, previous research [38][39][40][41] on LBSNs has also studied user's check-in data to predict user's location and mobility patterns. While [42][43][44] studied the uses and patterns of LBSN and examined the factors that predict the use of LBSNs regarding check-in.
For instance, mobile phone datasets have been used to understand the crowd and individual mobility patterns [45][46][47]. However, mobile phone data sets are not the only choice to study human mobility pattern analysis. Many other data sources of big data are collected and used, especially including geo-tagged data. This variety of new data sources is so diverse that it ranges log files from smart devices and websites, social media data and geo-tagged audio, video, and graphics data [48,49]. Ye, et al. [50] Proposed a novel definition of life pattern by presenting LP normal form to formalize the definition of individual life patterns and LP-Mine, an abstraction-and-mining framework to effectively retrieve life patterns from GPS data.
Many researchers [51][52][53][54][55] have concentrated on human mobility patterns, venue tagging, and check-in behavior toward using location-based social networks. Automatic venue tagging is one of the new concepts to observe spatial differences in many applications [56,57]. However, Gao and Liu [58] argued that when human mobility is integrated into an application that ranked locations based on a user's check-in history, temporal features were shown to be irrelevant. Ye, et al. [59] explored socio-spatial properties among different LBSN platforms, in another study Ye, et al. [60] analyzed check-in patterns of Foursquare users. A place to healthy relationships has been explored in [61] to expand opportunities for public health. Scellato, et al. [41] Presented a broad study of the spatial properties of the social networks arising among users in online location-based services and analyzed large dataset aimed to observe the inconsistency of urban spaces. Noulas, et al. [62] Explored user participation and provided insight of the city by analyzing social media data from foursquare in Seoul city and specially observed venues. Yu, et al. [63] Applied DBSCAN algorithm to observe Weibo locations in Shanghai and compared with k-means.
Location based datasets have now been used in many studies for urbanization and its environmental effects [64], development and prediction [65][66][67], travel and activity patterns [68,69] and emergency response [70][71][72] and urban sustainability [73]. Hong [74] Highlighted the use of an LBSN data to observe the willingness of buyers to pay for various factors. Visit frequencies can represent opinions and the geographical preferences of the individuals for places and given different motivations. Liu, et al. [75] Identified the factors that might cause the outbreak of Ebola and investigated the reaction by China, using big data analysis and explored differences in check-in behavior by gender. For example, Blumenstock, et al. [76] analyzed call detail record (CDR) data from Rwanda to observe population density and mobile phone use behavior by different genders. Wu, et al. [77] Highlighted the importance of big data as a tool to observe users' daily movement patterns and demographics specifically for housing prices. Preoţiuc-Pietro and Cohn [78] Studied the relationship between shared geo-locations and structured the nature of social connections. Kylasa, et al. [79] Introduced a new novel technique "activity correlation spectroscopy" for deriving connections by using the spectral and distributional structure of activity correlation within a set. Presently, there are some LBSNs available, including the focal ones in the present study. We infer that current research is helpful to understand gender differences and check-in behavior without considering gender equality.

Study Area and Data Source
In China, finding open and dependable data that describe geo location-based gender segregation is very hard. The LBSN dataset we are using in the current study comes from Chinese microblog Weibo during January-March 2016.
Shanghai, China (lying between 30 • 40 -31 • 53 N and 120 • 52 -122 • 12 E [80]) is located on the eastern edge of the Yangtze River Delta [81]. According to Gu, et al. [82] in 2015, Shanghai had a total area of 8359 km 2 , with a gross domestic product of 366 billion USD. The disposable income per capita of Shanghai is 7333 USD, where the income per capita of urban residents is 7788 USD and the income per capita of rural residents is 3412 USD [83]. As of 2015, the agricultural land area in Shanghai was 317,926 ha. The construction land area was 301,709.27 ha, and the unused land area was 193,564.46 ha. Shanghai is considered to be the most populated and dense community in the world (by urban area inhabitants), and a significant international center for trade, trade, tourism and fashion with a population of around 24.15 million people. In 2016, Shanghai is divided into 16 county-level divisions: 15 districts (Baoshan, Changning, Fengxian, Hongkou, Huangpu, Jiading, Jingan, Jinshan, Minhang, Pudong New Area, Putuo, Qingpu, Songjiang, Xuhui, and Yangpu) and 1 county (Chongming) [84]. Seven of the districts (Changning, Hongkou, Huangpu, Jingan, Putuo, Xuhui, and Yangpu) are located in Puxi (literally Huangpu West). These seven districts are referred as downtown Shanghai or the city center [85,86], as shown in Figure 1. al. [79] Introduced a new novel technique "activity correlation spectroscopy" for deriving connections by using the spectral and distributional structure of activity correlation within a set. Presently, there are some LBSNs available, including the focal ones in the present study. We infer that current research is helpful to understand gender differences and check-in behavior without considering gender equality.

Study Area and Data Source
In China, finding open and dependable data that describe geo location-based gender segregation is very hard. The LBSN dataset we are using in the current study comes from Chinese microblog Weibo during January-March 2016.
Shanghai, China (lying between 30°40′-31°53′ N and 120°52′-122°12′ E [80]) is located on the eastern edge of the Yangtze River Delta [81]. According to Gu, et al. [82] in 2015, Shanghai had a total area of 8359 km 2 , with a gross domestic product of 366 billion USD. The disposable income per capita of Shanghai is 7333 USD, where the income per capita of urban residents is 7788 USD and the income per capita of rural residents is 3412 USD [83]. As of 2015, the agricultural land area in Shanghai was 317,926 ha. The construction land area was 301,709.27 ha, and the unused land area was 193,564.46 ha. Shanghai is considered to be the most populated and dense community in the world (by urban area inhabitants), and a significant international center for trade, trade, tourism and fashion with a population of around 24.15 million people. In 2016, Shanghai is divided into 16 county-level divisions: 15 districts (Baoshan, Changning, Fengxian, Hongkou, Huangpu, Jiading, Jingan, Jinshan, Minhang, Pudong New Area, Putuo, Qingpu, Songjiang, Xuhui, and Yangpu) and 1 county (Chongming) [84]. Seven of the districts (Changning, Hongkou, Huangpu, Jingan, Putuo, Xuhui, and Yangpu) are located in Puxi (literally Huangpu West). These seven districts are referred as downtown Shanghai or the city center [85,86], as shown in Figure 1. In addition to the information available in Weibo dataset like user id, date, and time, we also have additional metadata like gender, geo-location (longitude and latitude), venue name, and category, but no personal information like the name is available. Therefore, check-in data records the In addition to the information available in Weibo dataset like user id, date, and time, we also have additional metadata like gender, geo-location (longitude and latitude), venue name, and category, but no personal information like the name is available. Therefore, check-in data records the daily life patterns and user's behaviors towards the services, and it reflects the average person's day-to-day operations. Table 1 describes the necessary information about Shanghai dataset.

Methodology
In this paper, we analyzed geo-location data that includes the user(s) ID, time, geo-coordinates (longitude and latitude), and the venue name and category. Figure 2 presents the process flow of data collection and check-in behavior analysis. daily life patterns and user's behaviors towards the services, and it reflects the average person's dayto-day operations. Table 1 describes the necessary information about Shanghai dataset.

Methodology
In this paper, we analyzed geo-location data that includes the user(s) ID, time, geo-coordinates (longitude and latitude), and the venue name and category. Figure 2 presents the process flow of data collection and check-in behavior analysis.  The frequency analytics methodology is divided into two stages: LBSN data collection and data analysis. The primary task of data collection phase is to download a large number of Weibo data in JavaScript Object Notation (JSON) format by using a python-based Weibo API as shown in Figure 2. However, in the data analysis stage, the critical task is to extract and analyze the feature of check-in data by considering location, time and gender. The analysis phase uses statistical and network analysis and data visualization to produce density maps and trends.
Weibo data is pre-processed to avoid noise and invalid records are filtered using the following criteria: a. Each check-in must have following information available: user id, date, time, gender, geolocation (longitude and latitude); b. The location of check-in is in Shanghai based on geo-coordinates as shown in Figure 1; c. The check-in lies within the date and time for the sampled data set; d. User(s) must have checked-in at least twice in a month, and the users with only one check-in record are considered invalid.
Before detecting hot-spots for check-in behavior, we analyzed check-ins by using a kernel density estimation (KDE) for estimating density function used in [79,[87][88][89] to produce a smooth density surface of check-in hot-spots in geographic space [90].
In our study, we considered the data available in the form of geo-tagged check-in. Let "C" be a set of historical check-in data i.e., C = {c 1 ,……, c n } where c i = <x,y> is a geo-location of the check-in 1 < i< n, of individual "i" and on time "t", where "C" is referred as the data set used.  Figure 3 presents a general framework for check-in frequency analytics. The frequency analytics methodology is divided into two stages: LBSN data collection and data analysis. The primary task of data collection phase is to download a large number of Weibo data in JavaScript Object Notation (JSON) format by using a python-based Weibo API as shown in Figure 2. However, in the data analysis stage, the critical task is to extract and analyze the feature of check-in data by considering location, time and gender. The analysis phase uses statistical and network analysis and data visualization to produce density maps and trends.
Weibo data is pre-processed to avoid noise and invalid records are filtered using the following criteria: a.
Each check-in must have following information available: user id, date, time, gender, geo-location (longitude and latitude); b.
The location of check-in is in Shanghai based on geo-coordinates as shown in Figure 1; c.
The check-in lies within the date and time for the sampled data set; d.
User(s) must have checked-in at least twice in a month, and the users with only one check-in record are considered invalid.
Before detecting hot-spots for check-in behavior, we analyzed check-ins by using a kernel density estimation (KDE) for estimating density function used in [79,[87][88][89] to produce a smooth density surface of check-in hot-spots in geographic space [90].
In our study, we considered the data available in the form of geo-tagged check-in. Let "C" be a set of historical check-in data i.e., C = {c 1 , . . . . . . , c n } where c i = <x,y> is a geo-location of the check-in 1 < i < n, of individual "i" and on time "t", where "C" is referred as the data set used.
where "c" refers to the location of check-in in training dataset "C" with bandwidth "h". It is assumed that the value of "h" is dependent on the resulting density estimate "f KD " which generates smooth density surface around "C" on data point "c i ." where "c" refers to the location of check-in in training dataset "C" with bandwidth "h". It is assumed that the value of "h" is dependent on the resulting density estimate "fKD" which generates smooth density surface around "C" on data point "c i ." Compared with the grid maps, kernel density estimation provides smooth distributions by eliminating the local noise to a certain degree by providing a non-parametric probability distribution with optimal bandwidth used to minimize the error. From the kernel density results, we reveal the dynamic of the city in both space and time in different days of the week in various districts of Shanghai.
We hope our results are useful for a behavioral study of users in regions by analyzing their check-in frequency. Through density maps and trend graphs, we can show the check-in frequency of LBSN users in different districts of Shanghai and their behavior of check-in during different hours of the day, weekdays, and weekends.

Results and Discussion
For our experiments, we utilized the Weibo check-in data set and used KDE to analyze the density of check-in data. The overall density of check-ins during January-March 2016 can be observed in Figure 4, and it can be observed that the center of the city has a high density of check-ins, which is a normal behavior for a big city due to easy accessibility of transport (i.e., subway) and living facilities (i.e., food, entertainment). Moreover, the high density of check-ins can be observed near the district borders of Baoshan, Changning, Minhang, Putuo, and Pudong New Area as compared to the center of these district. Compared with the grid maps, kernel density estimation provides smooth distributions by eliminating the local noise to a certain degree by providing a non-parametric probability distribution with optimal bandwidth used to minimize the error. From the kernel density results, we reveal the dynamic of the city in both space and time in different days of the week in various districts of Shanghai.
We hope our results are useful for a behavioral study of users in regions by analyzing their check-in frequency. Through density maps and trend graphs, we can show the check-in frequency of LBSN users in different districts of Shanghai and their behavior of check-in during different hours of the day, weekdays, and weekends.

Results and Discussion
For our experiments, we utilized the Weibo check-in data set and used KDE to analyze the density of check-in data. The overall density of check-ins during January-March 2016 can be observed in Figure 4, and it can be observed that the center of the city has a high density of check-ins, which is a normal behavior for a big city due to easy accessibility of transport (i.e., subway) and living facilities (i.e., food, entertainment). Moreover, the high density of check-ins can be observed near the district borders of Baoshan, Changning, Minhang, Putuo, and Pudong New Area as compared to the center of these district.  To investigate the check-in frequency and behavior, we analyzed the data regarding gender (male and female) in 10 districts of Shanghai. Figure 5a,b shows the overall weekly check-in trend; which depicts that female users prefer to use Weibo more as compared to male users during the whole week as well as during weekdays and weekends in all districts of Shanghai. It is also observed that check-in frequency increases during Saturday and Sunday. Moreover, Figure 6a,b shows the check-in density in Shanghai. It is observed that female users prefer to use Weibo as compared to male users and hence justifies the results of Figure 5. In Shanghai, to observe the check-in trends of both male and female users, it is essential to measure the check-in frequency during weekends and weekdays over a period. In Figure 7a  To investigate the check-in frequency and behavior, we analyzed the data regarding gender (male and female) in 10 districts of Shanghai. Figure 5a,b shows the overall weekly check-in trend; which depicts that female users prefer to use Weibo more as compared to male users during the whole week as well as during weekdays and weekends in all districts of Shanghai. It is also observed that check-in frequency increases during Saturday and Sunday. Moreover, Figure 6a,b shows the check-in density in Shanghai. It is observed that female users prefer to use Weibo as compared to male users and hence justifies the results of Figure 5. To investigate the check-in frequency and behavior, we analyzed the data regarding gender (male and female) in 10 districts of Shanghai. Figure 5a,b shows the overall weekly check-in trend; which depicts that female users prefer to use Weibo more as compared to male users during the whole week as well as during weekdays and weekends in all districts of Shanghai. It is also observed that check-in frequency increases during Saturday and Sunday. Moreover, Figure 6a,b shows the check-in density in Shanghai. It is observed that female users prefer to use Weibo as compared to male users and hence justifies the results of Figure 5. In Shanghai, to observe the check-in trends of both male and female users, it is essential to measure the check-in frequency during weekends and weekdays over a period. In Figure 7a    Data is analyzed to observe weekly check-in distribution by gender (male and female) and is presented in Figure 9. To our surprise, the difference of check-in behavior during Saturday and Sunday observed in Figure 8b is mainly due to change in check-in behavior by female users. Same check-in behavior can be observed from Figure 9 by the male users during Saturday and Sunday in Changning and Xuhui. However, from Figure 9 noticeable change in check-in behavior can be observed during Saturday as compared to Sunday in most of the districts, i.e., Baoshan, Hongkou, Huangpu, Jingan, Minhang, Putuo, and Xuhui. In Shanghai, to observe the check-in trends of both male and female users, it is essential to measure the check-in frequency during weekends and weekdays over a period. In Figure 7a  Data is analyzed to observe weekly check-in distribution by gender (male and female) and is presented in Figure 9. To our surprise, the difference of check-in behavior during Saturday and Sunday observed in Figure 8b is mainly due to change in check-in behavior by female users. Same check-in behavior can be observed from Figure 9 by the male users during Saturday and Sunday in   To observe the daily check-in trend in 10 districts of Shanghai, we analyzed the trend in a 24 h period. Figure 10a presents the daily check-in trend in 10 districts of Shanghai; high usage trend is observed during the morning (06:30 a.m.-09:30 a.m.), in Shanghai. It is also observed that the trend continues to rise till midnight after 23:00 pm for both male and female users as shown in Figure 10b,c.
To further observe the change in check-in behavior, we used kernel density estimation and visualized the density maps for 10 districts of Shanghai. Figure 11 reveal the dynamic of the districts in both space and time in 10 districts of Shanghai. It can be clearly observed that the city center has more check-in density as well as more density is observed near the district borders. Data is analyzed to observe weekly check-in distribution by gender (male and female) and is presented in Figure 9. To our surprise, the difference of check-in behavior during Saturday and Sunday observed in Figure 8b is mainly due to change in check-in behavior by female users. Same check-in behavior can be observed from Figure 9 by the male users during Saturday and Sunday in Changning and Xuhui. However, from Figure 9 noticeable change in check-in behavior can be observed during Saturday as compared to Sunday in most of the districts, i.e., Baoshan, Hongkou, Huangpu, Jingan, Minhang, Putuo, and Xuhui.  To observe the daily check-in trend in 10 districts of Shanghai, we analyzed the trend in a 24 h period. Figure 10a presents the daily check-in trend in 10 districts of Shanghai; high usage trend is observed during the morning (06:30 a.m.-09:30 a.m.), in Shanghai. It is also observed that the trend continues to rise till midnight after 23:00 pm for both male and female users as shown in Figure 10b,c.
To further observe the change in check-in behavior, we used kernel density estimation and visualized the density maps for 10 districts of Shanghai. Figure 11 reveal the dynamic of the districts in both space and time in 10 districts of Shanghai. It can be clearly observed that the city center has more check-in density as well as more density is observed near the district borders. The gender difference in 10 district of Shanghai is examined by the comparison of male and female users check-ins in 10 districts of Shanghai during January-March 2016. We use a relative difference [68,91] (dr) to calculate the gender differences in 10 districts of Shanghai, it is often used as a quantitative indicator of quality assurance and quality control in the proportion of all check-ins and is expressed as follows: where "Pm" and "Pf" denote the check-in probability of male and female users in 10 districts of Shanghai during January-March 2016. Gender differences in 10 districts of Shanghai are pragmatically explored at the cumulative level. First, we calculated the gender differences of in check-ins in 10 districts of Shanghai as a percentile of total accumulated check-ins made during January-March 2016. Table 2 displays the results of the relative difference calculated by using the Equation (3) during weekday and weekend. In Table 3, the relative difference values for the Saturday and Sunday are significantly larger than other days. Also, the relative difference values associated with Friday and Saturday are more than 0.55, while the values for the other days lies between 0.5. Results in Table 4 indicate that at the cumulative level, there are relatively significant gender differences in the number of check-ins in some districts (i.e., Huangpu, Pudong New Area, and Xuhui) by Weibo users in Shanghai. Results reveal that female users are more likely to use Weibo during the whole week, days and even in all 10 studied districts of Shanghai, whereas male users are apt to use Weibo during the weekday as compared to the weekend, as shown in Table 5. To further observe the change in check-in behavior, we used kernel density estimation and visualized the density maps for 10 districts of Shanghai. Figure 11 reveal the dynamic of the districts in both space and time in 10 districts of Shanghai. It can be clearly observed that the city center has more check-in density as well as more density is observed near the district borders.
The gender difference in 10 district of Shanghai is examined by the comparison of male and female users check-ins in 10 districts of Shanghai during January-March 2016. We use a relative difference [68,91] (d r ) to calculate the gender differences in 10 districts of Shanghai, it is often used as a quantitative indicator of quality assurance and quality control in the proportion of all check-ins and is expressed as follows: where "P m " and "P f " denote the check-in probability of male and female users in 10 districts of Shanghai during January-March 2016. Figure 11. Check-in densities in the 10 districts of Shanghai. Figure 11. Check-in densities in the 10 districts of Shanghai.
Gender differences in 10 districts of Shanghai are pragmatically explored at the cumulative level. First, we calculated the gender differences of in check-ins in 10 districts of Shanghai as a percentile of total accumulated check-ins made during January-March 2016. Table 2 displays the results of the relative difference calculated by using the Equation (3) during weekday and weekend. In Table 3, the relative difference values for the Saturday and Sunday are significantly larger than other days. Also, the relative difference values associated with Friday and Saturday are more than 0.55, while the values for the other days lies between 0.5. Results in Table 4 indicate that at the cumulative level, there are relatively significant gender differences in the number of check-ins in some districts (i.e., Huangpu, Pudong New Area, and Xuhui) by Weibo users in Shanghai. Results reveal that female users are more likely to use Weibo during the whole week, days and even in all 10 studied districts of Shanghai, whereas male users are apt to use Weibo during the weekday as compared to the weekend, as shown in Table 5.
Moreover, as observed from Figure 11, high values of check-ins are located in at the district boundaries, and the reason for this might be the significant proportion of financial and commercial activities. Finally, all the results imply that female users are more likely to use Weibo in 10 districts of Shanghai as compared to male users.

Conclusions
In the current study, we presented an in-depth empirical investigation of check-in behavior using intensity maps and trends using LBSN data. We investigated the check-in behavior from several different angles: the difference in gender, during weekdays and weekends, and daily and hourly patterns. In our results, we observe high rates of social media usage from female users and differences in check-in behavior during weekdays and weekends in all studied districts of Shanghai.
Apart from the inherent limitations of LBSN data, we discuss here to what extent LBSN data can be exploited to observer check-in behavior. More specifically, compared to other data sources (such as survey, census, GPS traces and call detail records), LBSN check-in data have some advantages, such as low cost and high spatial precision. However, check-in data also has some limitations, such as bias of gender, a low sampling frequency, and bias of location category. In summary, LBSN data is more likely to be a supplement to than a substitute of traditional data sources.
Based on the results of the empirical study, LBSN data has the potential to provide a new outlook as a supplement to observe gender differences and intensity of check-ins (during weekdays and weekends) and can help policymakers to define policies regarding the supply of services in urban areas within a city. It can also help to observe variations in population density over the period and act as a tool to estimate the supply of services in the city.
In the future, we plan to use LBSN data as a means to investigate the factors that influence the change in human check-in behavior within the city.
Author Contributions: Muhammad Rizwan, Wan Wanggen and Ofelia Cervantes conceived the research; Muhammad Rizwan designed the research, performed the simulations and wrote the article; Ofelia Cervantes and Luc Gwiazdzinski proof read the article for language editing. All authors read and approved the final manuscript.