A Study of User Activity Patterns and the Effect of Venue Types on City Dynamics Using Location-Based Social Network Data

Naimat-Ullah Khan, Wanggen Wan, Shui Yu, A. A. M. Muzahid, Sajid Khan, Li Hou

TL;DR

This paper leverages Weibo check-in data from Shanghai to analyze user activity patterns across different venue types. By combining temporal, spatial, and visualization techniques, it reveals how venue categories influence urban dynamics and user behavior, offering insights for smart city development and recommendation systems.

What Problem Does This Paper Address?

Understanding human behavior and urban dynamics is crucial for city planning and development. Traditional methods of collecting such data (surveys, interviews) are often time-consuming and may not capture the nuances of real-time activities. Location-Based Social Networks (LBSNs) offer a rich source of geo-tagged data, but analyzing this "big data" effectively, especially in diverse urban environments like Shanghai, presents challenges. Specifically, there's a need to go beyond general spatio-temporal analysis and understand how the type of location (venue category) influences user activity and contributes to the overall urban landscape. This paper aims to bridge this gap by classifying LBSN check-ins into distinct venue categories and analyzing their spatio-temporal patterns and density contributions in Shanghai.

Methods

The researchers employed a multi-faceted approach:

  • Data Collection & Preprocessing: Data was collected from Weibo, a popular Chinese LBSN, using its API. The raw data was then cleaned, filtered for relevant attributes, and transformed into a Geographical Information Systems (GIS) format.
  • Temporal Analysis: Descriptive statistics were used with SPSS to analyze check-in patterns over different time scales: hourly, daily, weekly, and the entire study period (one year). This helped identify typical human behavior related to work, leisure, and sleep cycles.
  • Venue Categorization: Venues were classified into 10 categories (e.g., Educational, Entertainment, Food, Residential, Shopping & Services) based on their names and geo-coordinates, comparing them to the function of physical locations.
  • Spatial Analysis: Kernel Density Estimation (KDE) was implemented using ArcGIS to visualize and quantify the spatial distribution and density of check-ins. This was done both for the overall city and for each venue category.
  • Visualization: Heatmaps, graphs, and tables were used extensively to present the findings, illustrating temporal trends, venue contributions, and spatial densities.

Datasets

  • Weibo Check-in Data: The primary dataset was sourced from Weibo, a Chinese microblogging platform. It comprised spatio-temporal check-in data for Shanghai over a one-year period (July 1, 2016, to June 30, 2017). The filtered dataset included 441,471 check-ins from 144,582 users across 20,171 venues.
  • OpenStreetMap (OSM): Map attributes from OSM were used to aid in the spatial analysis, providing context such as city boundaries, district boundaries, and metro lines.

Key Insights and Ideas

  • The "Invisible" City Life: The study highlights that even during late-night hours in a mega-city like Shanghai, there's a significant dip in social activity, likely due to sleep routines, suggesting a predictable rhythm to urban life.
  • Weekend vs. Weekday Dynamics: Saturdays see more check-ins than Sundays, possibly because people are less concerned about early mornings on the following day.
  • Migration's Impact: Major migration events, like the Chinese Spring Festival, significantly impact check-in frequency, causing a noticeable drop as people travel to their hometowns.
  • Venue Type Matters: Different venue types have distinct spatio-temporal footprints. "Residential," "Entertainment," and "Shopping & Services" venues attract the most check-ins, indicating their importance in daily life and leisure.
  • Density vs. Diversity: While city centers are generally denser, the diversity of check-ins can extend to suburban areas, influenced by factors like parks, residential complexes, and transportation hubs.

Key Results

  • Temporal Patterns: Daily analysis showed a clear pattern of low activity from midnight to early morning, increasing during working hours, peaking in the evening, and declining at night. Weekly analysis indicated higher activity on Saturdays compared to other weekdays.
  • Venue Category Contributions: "Residential" venues had the highest number of check-ins and users. "Hotel," "Food," and "Sports" categories had fewer check-ins. "Entertainment" and "Shopping & Services" venues showed less impact from summer vacations and festivals compared to other categories.
  • Spatial Density: High check-in densities were observed in Shanghai's city center (e.g., Hongkou, Huangpu, Jingan), with density decreasing as one moves towards suburban areas. However, specific areas with amenities like parks, residential apartments, and metro lines also showed high densities.
  • Category-Specific Densities: "Educational" and "Food" venues showed more concentrated densities, suggesting users visit specific, preferred locations. "Entertainment" and "Shopping & Services" venues, while popular, contributed to a more diverse density spread. "Residential" venues showed highly concentrated densities due to the large number of apartments.

Limitations and Future Work

  • Data Source Bias: The study acknowledges that relying solely on Weibo data might introduce bias, as it may not represent all user demographics or activities. Future work could involve integrating data from other LBSNs (e.g., WeChat, TripAdvisor) or different data sources to obtain a more comprehensive picture.
  • Data Completeness: Despite filtering, some data might be missing or irrelevant.
  • Cross-City Comparison: The authors suggest future research could compare datasets from different LBSNs or analyze similar patterns across multiple cities to explore cultural differences.
  • Applications: The findings are deemed beneficial for developing smart cities, recommendation systems, and LBSN studies in fields like transport, tourism, and food.

Summary updated

Abstract

The main purpose of this research is to study the effect of various types of venues on the density distribution of residents and model check-in data from a Location-Based Social Network for the city of Shanghai, China by using combination of multiple temporal, spatial and visualization techniques by classifying users’ check-ins into different venue categories. This article investigates the use of Weibo for big data analysis and its efficiency in various categories instead of manually collected datasets, by exploring the relation between time, frequency, place and category of check-in based on location characteristics and their contributions. The data used in this research was acquired from a famous Chinese microblogs called Weibo, which was preprocessed to get the most significant and relevant attributes for the current study and transformed into Geographical Information Systems format, analyzed and, finally, presented with the help of graphs, tables and heat maps. The Kernel Density Estimation was used for spatial analysis. The venue categorization was based on nature of the physical locations within the city by comparing the name of venue extracted from Weibo dataset with the function such as education for schools or shopping for malls and so on. The results of usage patterns from hours to days, venue categories and frequency distribution into these categories as well as the density of check-in within the Shanghai and contribution of each venue category in its diversity are thoroughly demonstrated, uncovering interesting spatio-temporal patterns including frequency and density of users from different venues at different time intervals, and significance of using geo-data from Weibo to study human behavior in variety of studies like education, tourism and city dynamics based on location-based social networks. Our findings uncover various aspects of activity patterns in human behavior, the significance of venue classes and its effects in Shanghai, which can be applied in pattern analysis, recommendation systems and other interactive applications for these classes.
BibTeX
@article{khan2020study,
  author = {Naimat-Ullah Khan and Wanggen Wan and Shui Yu and A. A. M. Muzahid and Sajid Khan and Li Hou},
  title = {{A Study of User Activity Patterns and the Effect of Venue Types on City Dynamics Using Location-Based Social Network Data}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2020},
  url = {https://openreview.net/forum?id=gKJw8I7bWH},
}
Figure