Abstract
Understanding urban travel behavior (TB) is critical for advancing urban transportation planning practice and scholarship; however, traditional survey data is expensive (because of labor costs) and error-prone. With advances in data collection techniques and data analytic approaches, urban big data (UBD) is currently generated at an unprecedented scale in relation to volume, variety, and speed, producing new possibilities for applying UBD for TB research. A review of more than 50 scholarly articles confirms the remarkable and expanding role of UBD in TB research and its advantages over traditional survey data. Using this body of published work, a typology is developed of four key types of UBD—social media, GPS log, mobile phone/location-based service, and smart card—focusing on the features and applications of each type in the context of TB research. This paper discusses in significant detail the opportunities and challenges in the use of UBD from three perspectives: conceptual, methodological, and political. The paper concludes with recommendations for researchers to develop data science knowledge and programming skills for analysis of UBD, for public and private sector agencies to cooperate on the collection and sharing of UBD, and for legislators to enforce data security and confidentiality. UBD offers both researchers and practitioners opportunities to capture urban phenomena and deepen knowledge about the TB of individuals.
Human movement constitutes the major dynamic of various spatial and temporal phenomena, such as human settlement, urban travel, freight transport, and the spread of contagious disease ( 1 ). Research about human movement and its relation to a broad spectrum of disciplines has progressed over the last five decades ( 2 ). Travel behavior (TB) (or human mobility) is consequently a critical topic in urban planning scholarship because (a) transportation planners seek to develop models to predict how people travel in time and space; (b) public transit agencies attempt to understand the factors that affect travel-related choices; and (c) policymakers expect to allocate transport resources more effectively and efficiently to encourage people to use public transit.
For decades, surveys have been the most commonly used method to obtain data for TB research. For example, decennial census information has been used to aggregate travel characteristics (e.g., commuting time or mode of transport) through random sampling, and regional or national travel surveys (e.g., National Household Travel Survey) collect self-reported data about users’ activities and travel via paper, Internet, or telephone interviews ( 1 , 3 , 4 ). While these types of data have contributed significantly to traditional TB studies, they are unavoidably labor-intensive, error-prone, and not cost-effective ( 1 , 5 ). Additionally, survey subjects are mostly recruited in a targeted fashion, resulting in actively solicited and small-sized datasets ( 6 ). However, the advent and pervasiveness of urban big data (UBD) offer an alternative for enriching available data for human mobility analysis.
UBD is rapidly evolving because cities, as complex systems, are major generators of big data in this digital era when new technological, institutional, social, and economic transformations occur rapidly ( 7 ). UBD is capable of capturing many important facets of cities and providing massive information about how individuals interact with urban environments over space and time ( 4 ). The prevalence of UBD also makes it possible for urban researchers to analyze individual TB, investigate spatial and temporal travel pattern changes, and understand, at a macro level, how people move within cities or between cities. It is therefore timely and important to examine the use of UBD and its various scholarly and professional applications in city planning.
The abundance of UBD has driven a new wave of TB research ( 1 ). Consequently, this paper explores four types of UBD in significant detail—social media data, GPS log data, mobile phone/location-based service (LBS) data, and smart card data—to study human mobility (leaving aside other research topics, such as public health). A typology is presented that captures the unique characteristics of each data type and explains its effectiveness in TB research. This paper also provides a critical discussion of opportunities associated with UBD, and challenges which are now faced or may be encountered in the future.
The remainder of this paper is organized as follows. The following section introduces the definition, categorization, features, and applications of UBD, and current TB research relevant to each category of UBD is reviewed. The subsequent section presents research opportunities and challenges associated with UBD, which aim to inspire scholars and practitioners to effectively conduct UBD-driven TB research. In the final section, the use of UBD in TB research is synthesized, and then the paper is concluded by offering new insights for planners and scholars to pursue in future studies.
Urban Big Data in Travel Behavior Research
Urban Big Data Defined
While there is no consensus concerning an accepted definition of UBD, the term generally refers to large dynamic and static datasets generated in an urban context from various urban infrastructures, organizations, and individuals ( 8 ). UBD usually contains information about urban phenomena, including physical elements (buildings, streets, environments), and people (and their characteristics and behaviors) and their relations (culture, public order). UBD provides ideal opportunities for studying individual TB.
Before the development of information and communications technology (ICT), which makes interactions between people and devices easier, survey data served as the major information source for human mobility analysis ( 1 ). However, researchers have criticized the use of surveys for urban transport studies. Unlike datasets derived from infrequent surveys or interviews with high acquisition cost, UBD overcomes these weaknesses because ICT facilitates UBD generation and collection ( 2 , 9 ).
UBD, generally speaking, can yield richer individual spatial-temporal information for TB research (compared with traditional survey data) for three reasons. First, UBD is generated from multiple sources at unprecedented scales in relation to volume, variety, and speed. Second, UBD can fill the gap where survey data is limited or nonexistent. Third, UBD provides massive person-level data with spatial-temporal coverage, creating possibilities to examine travel preferences comprehensively.
UBD evolves from big data and possesses three basic “v” characteristics: volume, velocity, and variety ( 4 , 10 ). (See Figure 1a). Volume suggests the massive amount of data, often in terabytes or petabytes, making it larger in size and richer in detail than traditional survey data. Velocity refers to continuous and real-time data generation processes. Variety indicates diverse types and formats of UBD collected from wide sources. For example, massive social media data generated from mobile device applications can be obtained anytime and anywhere with few geographical constraints. A fourth “v,”value, was added later to capture the contribution of big data to research and business ( 11 ). (See Figure 1b).
Interrelated features of urban big data: (a) left, (b) middle, and (c) right.
Pan et al. highlighted three facets—hierarchy, integrity, and correlation—in addition to the four abovementioned “v” characteristics to reach a more comprehensive picture of UBD, as depicted in Figure 1c ( 8 ). Hierarchy refers to the organization of individual-level data in nested levels, reflecting the structure of physical and social systems within cities. For example, individual mobile phone application usage data can be hierarchically summarized by geographical scales (e.g., municipalities), social systems (e.g., minority communities), or physical systems (e.g., public agencies or government) for various purposes. Integrity refers to the assurance of accuracy and consistency, as well as the broad coverage of complexity in urban environments ( 8 , 12 ). This relates to veracity, which describes the reliability and accuracy of data. Correlation denotes the connections between various types of UBD.
Typology of UBD and Applications
UBD can be classified from various dimensions, such as data usage, sources, or application objectives. One well-recognized method to categorize UBD is based on the user community associated with sources: (a) sensor data from urban infrastructure and moving objects (e.g., data from GPS-enabled devices); (b) user-generated information (e.g., social media data, web-searching data); (c) governmental administration data (e.g., confidential micro-data); (d) customer and transaction records from private business or independent developers; (e) aspects of arts and humanities (e.g., film, sound recording, or other media); and (f) integrated data ( 5 ). It is worth noting that these classifications are not mutually exclusive; for instance, sensor systems for collecting UBD might be owned by governmental agencies or private companies ( 5 ).
This study focuses primarily on four commonly used types of UBD for TB research: social media data, GPS log data, mobile phone/LBS data, and smart card data. The first type is related to user-generated information. The second and third are usually obtained through sensors embedded on vehicles or urban infrastructure. The last type, smart card data, may derive from either governmental administration data or customer records from the private sector, depending on the ownership of the operating public transit agency.
To perform this research, an extensive literature search was conducted using relevant keywords, including “(urban) big data,”“travel behavior,” or “human mobility.” The bibliographies of scholarly literature reviews of related topics were also examined to identify other pertinent studies, such as literature reviews of transportation and mobility, summaries of big data and urban informatics-related conference papers, and research about big data applications and urban studies in China ( 5 , 8 , 12 ). More than 50 scholarly articles published in peer-reviewed outlets were identified addressing various aspects of UBD; some articles were excluded from further consideration because of unrelated research aims (e.g., using UBD to study epidemiology or event monitoring).
The authors acknowledge several limitations of this study. First, given the vast amount of literature on this topic, the 50+ articles reviewed provide at the moment a sufficient amount of relevant literature to explain UBD and distill associated challenges. As new research is published, however, the authors urge scholars to continue to cultivate a comprehensive understanding of UBD in TB research. Second, as UBD continues to develop, new challenges loom around the corner. Three aspects of these challenges are identified later, while, in reality, there are potentially other technological and theoretical concerns and unpredicted problems associated with the use of UBD awaiting solutions.
As the chief outcome of this in-depth literature review, the 26 most relevant and representative empirical studies are presented, categorized into a typology (Table 1).
Travel Behavior Studies Categorized by Four Types of Urban Big Data (UBD)
Social Media Data
Social media was intended to facilitate people sharing their daily activities and other aspects of contemporary life. Through social network applications, such as Facebook, Twitter, Foursquare, and Flickr, people can share personal experiences and locations by posting comments, photos, or check-in status. As social media data is usually geotagged and time-stamped, it has unique social and spatial properties that are extremely useful for TB studies. Social media data also enables researchers to analyze the content or semantic meaning of a message or a photo and to investigate travel patterns of individuals at fine resolution ( 1 , 38 ). Deep learning analysis has been applied to social image data (e.g., Flickr photos), providing an alternative to discover tourists’ behaviors and understanding their tourism preferences ( 20 ).
Check-in records, which employ LBS data, are usually reported by users in social media applications and contain the time and location (at city or neighborhood level, or a specific building) of a visit, and other messages proactively shared by data contributors (e.g., emotions or comments). Liu et al. extracted inter-urban movements from check-in records and analyzed underlying spatial trip patterns and spatial interactions ( 13 ). Zhang et al. mapped individuals’ space-time footprints using location-based social media (LBSM) data (i.e., contents from Weibo) ( 15 ). Xing et al. also extracted hidden dynamic human activity from geotagged Tweets to construct a mixed land use evaluation model ( 16 ). Similarly, another study proposed a new framework to help enhance crime prediction accuracy with data from multiple sources, including Foursquare check-in reports, census data, and subway usage records ( 18 ). Cui et al. used Weibo check-in records (combined with survey data) to identify personal activity-specific places in Beijing, finding that integrating multiple sources of data can offer larger sample sizes, real-time updating, higher accuracy, and a lower acquisition cost ( 17 ).
Social media, which provides users a virtual environment for engagement with fewer constraints and more freedom to share their daily activities, is available at a micro-level and therefore facilitates TB research related to travel mode preferences or user experience. Schweitzer used Tweets to ascertain people’s perception of public transit service and demonstrated the potential of social media information to help transit agencies improve public transit operational efficiency ( 39 ). Agryzkov et al. compared Foursquare geo-located data with field-collected data to understand social activities taking place in Murcia, Spain, highlighting similarities in activity preferences between what was interpreted from social network users and what was obtained from field records ( 40 ).
Taken together, a growing body of published work uses social media data from various applications at multiple scales to investigate travel patterns. However, it cannot be neglected that the primary users of social media applications are the younger generation. Digitally excluded individuals, including low-income persons and older adults, may lack access to the Internet or mobile devices and are consequently under-represented in information gleaned from this data source. Therefore, social media data is not representative of the full population. Moreover, even if social media data is collected in large volumes, the portion of useable and reliable data is quite low, as the majority might be “trash data.”
GPS Log Data
Technological advancements in GPS-enabled devices have reinforced the availability of urban trajectory data for geographers and transportation planners. Trajectory data, also known as GPS-derived movement data, is potentially much richer than traditional movement data temporally and spatially because it is recorded at discrete time intervals associated with a chronological series of location points ( 26 ). GPS log data forms the basis of TB studies and usually contains longitude, latitude, altitude, direction, trip time, and travel speed; it can also be combined with other data, such as biological information, to respond to specific research needs ( 1 ).
There are two widely-used methods to collect GPS log data for TB research. First, log data is collected through self-carrying GPS devices. The first regional travel survey was conducted in Austin, Texas, in 1997, in which 117 households were equipped with GPS loggers to collect their trajectory records ( 41 ). One study explored human mobility patterns using GPS traces of 258 volunteers and verified the “Levy flight characteristic” (or super-diffusive) nature of human mobility ( 22 ). Another study similarly concluded that the heavy-tail flight distribution of human mobility induced the super-diffusivity of travel based on GPS traces collected from 101 volunteers in five sites ( 42 ). Smartphones can also serve as GPS trackers to record users’ locations and corresponding time, making it possible to infer transportation modes and travel sequences ( 21 ). Furthermore, a method was developed using GPS log data to detect abnormal behavior patterns among the elderly who may be suffering from early-stage Alzheimer’s disease ( 23 ).
Second, GPS receivers installed on vehicles can also record log data. Taxi travel trajectory data, for instance, can be recorded through installed GPS receivers, especially in urban areas ( 25 , 43 ). Liang et al. analyzed 20 million trajectories collected from 10,000 taxis in Beijing and found taxis’ traveling displacements in urban space tended to follow an exponential distribution ( 44 ). Another study combined GPS log data with social media check-in data to comprehensively examine travel patterns ( 24 ). These studies attempt to explore the relationship between urban spatial structure and human movement patterns, since GPS data has the flexibility and potential to illuminate what traditional travel surveys may not reveal ( 45 ).
Certain researchers focused on developing independent and interactive interfaces or applications for transportation planners to monitor real-time traffic status and take immediate necessary actions. Al-Dohuki et al. developed a new system called “SemanticTraj” for managing and visualizing taxi GPS trajectory data and potentially identifying hidden congestion spots ( 46 ).
The main limitation of GPS log data is that its accuracy is significantly affected by surrounding urban environments—including characteristics of urban form and building configuration—because GPS use is restricted in the open air and sensitive to urban canyon effects ( 1 ). Additionally, log data sometimes lacks demographic information about data providers, which may be resolved by performing post-event surveys of GPS device users (or implementing the simultaneous collection of both surveys and log data).
Mobile Phone/LBS Data
Mobile phone/LBS data applied to TB research generally emanate from two broad sources: (1) cellular networks and (2) mobile phone users ( 1 , 47 ). Cellular network-based data is collected by telecom companies, while mobile-phone-based data is contributed by phone or application users, usually collected by proactive researchers for academic or marketing purposes. It has been confirmed that cellular phones are suitable for tracking and monitoring individual travel patterns in real-world environments ( 48 ).
Call detail records (CDR) data and location-based information are two widely applied types of cellular network-based data in TB research. CDR data—usually including user ID, cell tower ID, and phone call duration—has been proven to be efficient to extract origin–destination (OD) trips, detect activities and trip types, identify spatial and temporal activity patterns, and infer travel modes ( 1 , 31 , 47 , 49 ). Jiang et al. introduced a series of algorithms to infer anonymous individuals’ potential stay locations and their trip purposes by linking information extracted from CDR to surrounding land uses of (potential) stays ( 49 ). They also validated the predictability of human activity dynamics using CDR data.
Location-based information or LBS data is typically determined using cellular towers, assisted-GPS, Wi-Fi, or other positioning technologies. Compared with CDR data, LBS data typically has higher and more stable temporal granularity, as well as finer spatial resolution ( 47 ). For example, Song et al. used real-time locating-request records from map applications on mobile phones—paired with night-time light data and satellite images—to quantify and assess dynamic changes in citizens’ exposure to greenspace ( 29 ). The study suggested that integration of multi-geospatial big data can increase assessment rates.
A great deal of intelligence can also be extracted from mobile-phone-based data to support TB research. Unlike social media data which is collected from social network applications via mobile phones and other devices, mobile-phone-based data is collected either through public applications developed by third-parties (to provide navigation services or LBS) or through dedicated applications designed by researchers for specific purposes ( 47 ). A new method was proposed to investigate driving offense—mobile phone use while driving—on the basis of driver exposure and behavior metrics using data collected using smartphone sensors ( 32 ). Another study examined GPS-enabled mobile phone location data volunteered by 127 university students to assess urban street connectivity and confirmed that such data could provide an accessible and flexible evidence base to support the design of urban spaces ( 30 ).
There has been abundant TB research using mobile phone data. Some scholars have conducted comprehensive literature reviews and attempted to establish a conceptual framework for TB studies related to UBD ( 1 , 47 , 49 ). Li et al. proposed a framework that combined both cell phone signal data and taxi GPS data from two large metropolitan areas in China to evaluate the relationship between mobility and urban built environments ( 24 ). Yuan et al. explored the correlation between the usage of mobile phones and individual TB through three indicators: travel radius, eccentricity, and entropy ( 27 ). Another study also applied correlation analysis for cell phone signal data and explored potential factors that related to the service radius of urban parks, with a particular interest in park preference ( 50 ). These studies have paved the road for the use of large-scale mobile phone data for conceptual TB studies, establishing frameworks for future research.
The unique features of mobile phone LBS data are their prevalent scale, and longitudinal and individual micro-level details. Unlike social media data with potential sampling bias, mobile phone data can generate overwhelming sampling penetration of the entire population, which is rare or impossible in previous TB studies using traditional survey data ( 1 ). However, concerns about privacy and illegal usage, as well as positioning resolution, still exist among mobile phone data providers.
Smart Card Data
Smart cards were originally designed for automatic fare collection on public transit systems ( 1 ). Some smart cards can even be utilized in shops, restaurants, or parking lots, such as the Octopus card (Hong Kong) and the Oyster card (London). Smart cards normally store value and transaction information (time, fare amount), trip details (travel mode, origin and destination, time, fare), and users’ personal information.
Smart cards can improve the quality of TB studies in several respects. First, from smart card data, researchers can extract exhaustive information at the individual level, which is impossible from onboard ticket sales ( 33 ). Second, it is feasible to identify round-trip travel and eliminate potential double-counting of trips, as many commuters will use the same mode on the same day. Third, it is practicable to measure the spatial variability of public transit use (frequency of use of bus stops for boarding) and temporal variability (proportion of zero-boarding days), which can guide transit agencies to better allocate resources to meet travel demand ( 34 ).
Bagchi and White first introduced a conceptual system of smart card data usage for TB research and demonstrated that smart cards could generate larger volumes of data over a longer period of time compared with existing data sources ( 33 ). Roth et al. used an Oyster card database of individual movements in the London Underground to depict the structure of the city and identify activity sub-centers ( 35 ). A web-based application called BusViz was proposed to use passenger fare card data and large streams of sensor data to monitor and visualize the ridership dynamics of the bus system in Singapore. This will not only capture public transit usage but also highlight regional movement patterns visible to bus service operators for better public resource allocation ( 36 ). More recent research efforts have combined smart card information with other types of data, such as vehicle GPS data, to conduct more comprehensive and scalable TB studies ( 37 ).
These empirical studies demonstrate that smart card data is reliable and valuable in TB and transport planning studies. However, certain limitations are acknowledged. First, only major cities worldwide, where public transit systems are maturely developed, such as Hong Kong, Singapore, and London, may adopt smart card systems and collect data suitable for research ( 1 ). Also, a traveler’s trip purpose and multimodal transfers are not easily detectable from smart card data. Lastly, the concern about privacy, a common issue related to UBD, is raised when linking trip details with personal information.
Discussion: Challenges and Opportunities
The technological disruption caused by UBD has brought new opportunities and approaches to urban research. As depicted in Table 1, certain analytical research methods from other disciplines have been proven to be quite useful in dealing with UBD. For example, data mining technology, although not a traditional analytical method directly used in urban planning studies, can effectively extract trips and user characteristics from social media data, mobile phone data, and GPS log data. Moreover, the advancement of technologies and abundance of UBD reinforce commonly used methods (including descriptive analysis and predictive analysis) for studying TB. Descriptive analysis of TB research using UBD focuses on revealing patterns, detecting clusters, and enhancing visualization. For example, visualization of travel patterns using LBS data from map navigation apps, visualization of LBSM data, and visualization of urban activity through mobile phone data all take advantage of massive amounts and variety of data to detect and reveal underlying patterns ( 15 , 29 , 51 ). The predictive analysis focuses on inference. For example, social network apps check-in data is used to infer urban clusters and predict crime, and GPS log data for older adults is used to increase travel mode prediction performance ( 18 , 21 ). The scholarly contribution of these studies is enhanced by the emergence and an abundance of UBD.
TB research benefits from new opportunities brought by UBD, but researchers are also facing new challenges. Among various types of challenges previously identified for applying UBD to urban research, next, two are chosen (methodological and political) and one is added (conceptual) for further elaboration ( 5 ).
Conceptual Challenges
Conceptual challenges pertain to two key aspects of UBD: (a) the innate limits and the bias of UBD, and (b) a lack of meaningful research questions and theoretical foundations. For example, social media data is often criticized for not being representative of broad populations and thus being unsuitable for scholarly research, as social media users are more likely to be younger, more digitally savvy, and geographically constrained (concentrated in certain places or at certain events) ( 52 , 53 ). Also, social media data mostly lacks demographic information of users, although merging multiple data sources, such as smart card data or census data, may serve as a solution to better understand the characteristics of users and their TB. Furthermore, social media content data is criticized for only providing subjective contextual particularities or local experiences; therefore, studies using such data lack generalizability ( 54 ). These innate flaws of user-generated UBD are challenging to avoid and still awaiting resolution from either methodological breakthroughs or technological leaps. Even if it is practicable to extract useful information from mobile phone data or GPS log data, several assumptions must be made before analyzing the data. As TB research using UBD continually “borrows” technologies, methods, and even theories from other disciplines to help urban researchers forecast demand for urban resources, simulate scenarios, and evaluate urban policies, this body of research is weakly undergirded by a theoretical foundation ( 5 ).
Another limitation for the use of UBD is scholars’ ability to ask practical research questions aimed at solving pressing challenges. Currently, many scholars are more enthusiastic about the possibilities for complex data analysis using UBD than they are about generating meaningful and relevant research questions. Thakuriah et al. claimed that, among current urban research using UBD, an inadequate focus was placed on the “why” or “how” of urban dynamics in lieu of simple cause-and-effect relationships ( 5 ). The essence of utilizing UBD in TB research should be discovering, revealing, and understanding patterns that cannot be explored through traditional data sets. Hao et al. also suggested that the role of UBD should be “leading innovation” and even causing an evolutional change in urban research ( 55 ).
Methodological Challenges
The methodological challenges are considered from two perspectives: collection methods and analytical methods. ICT indeed enhances the movement toward a data-rich environment, providing more options to collect diverse urban data. It is challenging, though, to collect, retrieve, and process information from a large volume of unstructured data ( 5 ). When retrieving and gathering social media data from a raw data stream, for instance, special data documentation is usually required, since pre-processing and storage of data occur simultaneously. This unique workflow needs further refinement, since it challenges existing data gathering practices.
Analyzing UBD requires researchers to acquire analytical skills and tools ( 28 ). However, urban planning education historically relies on using census data rather than UBD, making it challenging for any researcher to master the required skills ( 56 ). Some analytical methods, such as Latent Dirichlet Allocation (LDA), are not generally used by urban researchers; however, they can be effectively used to analyze mobile phone data, text messages from social media, and other unstructured data (see Table 1). Since UBD collection and analysis are tightly coupled, it is necessary for urban researchers to learn new skills or to have closer collaboration with experts from relevant disciplines, including data science, computer science, statistics, geography, and engineering ( 5 ).
Political Challenges
Political challenges related to using UBD stem from both external and internal factors. Internal factors, reflecting the inherent complexity of UBD, involve two aspects: (a) data access and (b) user privacy and data security. External factors comprise the actions of institutions and stakeholders ( 5 ).
Currently, vast collections of UBD are locked away because various economic, legal, and procedural limitations restrict access and use ( 5 ). A majority of UBD is owned by individual agencies or private organizations and firms (e.g., social media applications firms), with limited access for most scholars. When data is privately owned, researchers are either not permitted to purchase or unable to afford the cost of data acquisition. Data availability is also dictated by the terms of service agreements from private companies or constrained because of considerations toward maintaining competitive advantages; even for some government agencies, there are still highly restricted limitations on data use. Although some data can be accessed by API or other platforms, there are still limits on data access and downloads.
As discussed earlier, urban researchers often struggle with UBD acquisition. Scholars may be able to conduct more TB research if restrictions could be lifted. If policies could be made more user-friendly, confidential use of UBD could be assured. Cooperation between research institutions, government agencies, and the private and public sectors should be reinforced, since there will undoubtedly be more UBD generated in the future. Mutually beneficial agreements between public transit agencies and institutions, for instance, may lead to rapid data collection, easy data sharing, sophisticated but efficient data analysis, and real-world application. Additionally, the value of UBD will increase, like any public goods; publicly funded data collection infrastructure (e.g., kiosks in NYC) will also gain more value and consequently generate more UBD for research applications ( 5 ).
Lastly, privacy is another concern in employing UBD for TB research, especially projects using social media data. Privacy has been identified as a fundamental human right ( 57 ). UBD often provides detailed information at the individual level (e.g., time-stamped location, daily activities, and personal comments). It is consequently vital to guarantee that data providers can securely share data, that data will be analyzed for research purposes only, and that data will be protected. However, people may trade some of their privacy for equivalent benefits, such as other utilization of information or minimization of potential risks ( 58 ). To address this issue, Thakuriah et al. argued for intervention from state or federal agencies to use legislation as a lever to ensure security and confidentiality of the use of UBD ( 5 ). Such legislation will not only raise awareness among UBD providers and users but also encourage new technological approaches to guarantee privacy.
Conclusion
During recent decades, new ICT and data collection technologies have resulted in the emergence of UBD that are suitable for TB research. UBD offers both researchers and practitioners opportunities for using immense data sets capturing urban phenomena to deepen knowledge about individual TB and its larger effects on society. This review and synthesis can deepen scholars’ understanding of UBD in TB research. By identifying four types of UBD commonly used in TB research, it is found that social media data, GPS log data, mobile phone data, and smart card data all possess unique features that traditional survey data do not own, and each type of UBD has offered breakthroughs and accelerated the development of TB modeling and theoretical frameworks for the urban planning field. It is also found to be evident that UBD can be practical and useful when examining “hypotheses previously considered as missions impossible” using traditional datasets ( 1 ).
This review and synthesis also confirm that all four types of UBD offer massive spatial and temporal information at the individual level, which is ideal for analyzing individual TB or citywide travel patterns. For example, social media data is relatively easy to access with a low acquisition cost. It has been repeatedly suggested that combinations of UBD from various sources will improve the quality of integrated data and increase the validity of results. For example, GPS log data can yield a more comprehensive understanding of individual mobility patterns when combined with users’ demographic information from surveys, real-time social media data from social network applications, or travel data from local public transit agencies.
Although some limitations have been acknowledged, it is still worth noting that other forms of UBD exist but are not explored in this study. For example, certain data types have the power to contribute to TB research in novel ways—including moving object image data from vision-driven systems embedded in urban infrastructure, vehicle license plate image data, and automatic pedestrian counts collected from Laser Radar—if they are appropriately processed using artificial intelligence or machine learning technologies ( 59 ). These new technologies and under-explored data sources on the horizon create demands for planners and practitioners to pursue enhanced uses for UBD in the advancement of TB research.
To maximize the potential of UBD in TB research, the authors urge urban researchers to be aware of conceptual challenges, to deepen the theoretical foundations of scholarly research, and to develop cutting-edge research questions that explore meaningful topics. Scholars and practitioners should also equip themselves with additional analytical skills or reinforce collaboration with experts from relevant disciplines such as geography, computer science, and civil engineering. The authors also conclude that cooperation between private sectors and public agencies should be reinforced, and legislation can be used as a tool to ensure the secure and confidential use of UBD.
Footnotes
Author Contributions
The authors confirm contribution to the paper as follows: study conception and design: C. Wang; data collection: C. Wang; analysis and interpretation of results: C. Wang, D. B. Hess; draft manuscript preparation: C. Wang, D. B. Hess. All authors reviewed the results and approved the final version of the manuscript.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.