Source-linked AI summary
Combining smart card data and household travel survey to analyze jobs-housing relationships in Beijing
Ying Long, Jean-Claude Thill
TL;DR
The paper addresses limited smart-card-based analysis of jobs-housing relationships and commuting, alongside the lack of social information in location-based-service data. It combines bus smart card data, a household travel survey, and parcel-level land use in Beijing to identify locations and commuting routes, obtaining validated identification results and identifying locations for over one million cardholders. The approach is preliminary and is limited because smart card data cover only bus riders.
Problem
Smart card data have been used less for jobs-housing and commuting analysis, while location-based-service data lack rich social information for in-depth application.
Method
The paper combines one week of bus smart card data with a one-day household travel survey and parcel-level land use to identify jobs-housing locations and commuting routes in Beijing.
Results
Housing locations were identified for 1,045,785 cardholders, job locations for 362,882, and both locations for 237,223 cardholders.
Takeaways & Limitations
The experiment provides a preliminary step toward enriching location-based-service data by combining smart card, survey, and urban GIS data.
Takeaways & Limitations
Smart card data are limited to bus riders, excluding trips made by other modes.
Abstract
from arXiv · showhide
Location Based Services (LBS) provide a new perspective for spatiotemporally analyzing dynamic urban systems. Research has investigated urban dynamics using GSM (Global System for Mobile Communications), GPS (Global Positioning System), SNS (Social Networking Services) and Wi-Fi techniques. However, less attention has been paid to the analysis of urban structure (especially commuting pattern) using smart card data (SCD), which are widely available in most cities. Additionally, ubiquitous LBS data, although providing rich spatial and temporal information, lacks rich information on the social dimension, which limits its in-depth application. To bridge this gap, this paper combines bus SCD for a one-week period with a one-day household travel survey, as well as a parcel-level land use map to identify job-housing locations and commuting trip routes in Beijing. Two data forms (TRIP and PTD) are proposed, with PTD used for jobs-housing identification and TRIP used for commuting trip route identification. The results of the identification are aggregated in the bus stop and traffic analysis zone (TAZ) scales, respectively. Particularly, commuting trips from three typical residential communities to six main business zones are mapped and compared to analyze commuting patterns in Beijing. The identified commuting trips are validated on three levels by comparison with those from the survey in terms of commuting time and distance, and the positive validation results prove the applicability of our approach. Our experiment, as a first step toward enriching LBS data using conventional survey and urban GIS data, can obtain solid identification results based on rules extracted from existing surveys or censuses.
1 Introduction
The paper addresses limited use of smart card data for jobs-housing and commuting analysis by applying it to urban spatial analysis in Beijing. It positions smart card data as a continuous, large-scale complement to conventional surveys and validates identified patterns with household travel data.
- Research gap: Large-scale microdata acquisition remains limited because urban-structure research still relies heavily on physical-space data, questionnaires, surveys, or censuses.The paper contrasts the limited availability of microdata with the growing use of location-based technologies.
- Smart card data: Smart card systems continuously record detailed bus transactions and can provide a complete, real-time travel diary for bus travelers.The paper also notes that smart card data can help validate traditional public-transit travel models.
- Smart card data: Conventional household travel surveys are expensive and infrequent, whereas smart card data offer intensive travel-pattern records at much larger scale.Smart card data are presented as a potential substitute for, or complement to, household travel surveys.
- Research gap: Location-based services have supported urban studies using GSM, GPS, SNS, and Wi-Fi, but smart card data have received less attention for jobs-housing and commuting analysis.The paper identifies this underexplored application as its main focus.
- Study contribution: Using Beijing as a case study, the paper identifies jobs-housing locations, their relationships, and commuting trips from smart card data, then validates the results with household travel survey data.The study presents jobs-housing and commuting analysis as a showcase application of smart card data to urban spatial analysis.
2 Jobs-housing trips in conventional household travel surveys
Conventional household travel surveys directly record trip purposes, locations, timing, modes, and traveler attributes, but their spatial resolution and coverage are constrained. The paper therefore considers smart card data alongside surveys to obtain larger-sample, finer-scale commuting information while supplying missing social and trip-purpose information through the survey.
- Survey information: Household travel surveys record socioeconomic attributes, trip origins and destinations, time, duration, purpose, and mode.Housing and job locations are directly recorded, usually at the traffic analysis zone scale.
- Survey limitations: Survey-based housing and job locations and trips are generally represented between traffic analysis zones rather than at fine spatial scales.Residential communities and business zones may be smaller than a traffic analysis zone.
- Survey limitations: Only a small portion of households in a city are surveyed because of time and cost constraints.This limits the coverage of conventional household travel surveys.
- Complementary data: Mining smart card data can provide more precise spatial resolution and a larger sample than household travel surveys, but cannot directly provide jobs-housing locations or commuting trips.The paper treats this combination of strengths and weaknesses as the basis for integrating both data sources.
- Analytical opportunity: Fine-scale identification can visualize commuting patterns from typical residential communities to business zones and potentially reveal additional jobs-housing information.The paper emphasizes that these locations are generally smaller than one traffic analysis zone.
- Complementary data: Smart card data lack cardholder socioeconomic attributes and trip purposes, while household surveys can supply this additional information for jobs-housing analysis.The paper describes combining the two sources as a promising approach.
3 Data
The study combines Beijing bus smart card data, household travel survey data, and parcel-level land-use data to characterize transit activity and support jobs-housing analysis. The datasets cover extensive bus trips, mapped transit infrastructure, land-use parcels, and detailed household travel diaries.
- Transit network: Beijing’s bus network contains 1,287 routes and 8,691 stops across the 16,410-km² Beijing Metropolitan Area.Opposite-side bus platforms are merged into one stop feature, and the average distance to the nearest neighboring stop is 231 m.
- Spatial units: The analysis uses 1,118 traffic analysis zones (TAZs), defined from administrative boundaries, main roads, and the planning layout.TAZ-level results exclude trips outside the BMA because no TAZs are available there.
- Land use: Parcel-level land-use data classify residential parcels as housing locations and commercial, public-facility, and industrial parcels as job locations.The dataset contains 133,503 parcels, including 29,112 residential parcels and 57,285 job parcels.
- Smart card data: The one-week 2008 SCD covers 77,976,010 bus trips made by 8,549,072 cardholders and excludes subway records.Each cardholder makes an average of 1.30 bus trips per day, and the cards are anonymous.
- Smart card data: Fixed-fare routes provide only departure time and stop ID, whereas distance-fare routes provide complete trip information.Incomplete fixed-fare records can prevent housing or job identification, although some transfer cases remain identifiable.
4 Approach
The approach preprocesses smart card records into daily bus travel diaries and two complementary representations: TRIP for rides and PTD for activity durations. Survey-derived rules and parcel land use then identify housing and job locations from one-day and one-week patterns.
- 4.1 Data pre-processing and data forms: Bus smart card records are geocoded to bus stops and combined into a full bus travel diary for each cardholder and day.The diary supports calculating daily trip counts, total bus travel time, total bus travel distance, and daily start and end points.
- 4.1 Data pre-processing and data forms: TRIP represents each bus ride as departure location and time plus arrival stop and time: TRIP = {OP, OT, DP, DT}.It is the direct expression of the bus travel diary and is used for commuting trip route identification.
- 4.1 Data pre-processing and data forms: PTD represents an activity at a bus stop with its location, start time, and duration: PTD = {P, t, D}.PTD is converted from TRIP, better matches time geography, and is used to identify urban activities such as housing and jobs.
- 4.2 Identification of housing and job locations using one-day data: Housing locations are initially assigned near the departure stop of the first daily trip, using an estimated average walking distance of 750 m.The rule is supported by the survey finding that 99.5% of first trips started from home.
- 4.3 Identification of housing and job locations using one-week data: One-week housing and job locations are selected with rule-based decision procedures that combine one-day location frequency, spatial clustering, and land-use-based residential or job potential.Locations within 500 m are grouped into one cluster, and the cluster with the greatest relevant potential can determine the final location.
5 Results
The Beijing analysis identifies jobs, housing, and commuting trips from smart card data, aggregates results spatially, and compares commuting patterns across areas and communities. Validation against survey data supports the applicability of the approach while revealing spatial clustering and uneven commuting flows.
- 5 Results: The workflow preprocesses smart card data with survey and GIS layers, then uses Python geoprocessing to identify locations, trips, and commuting patterns.The processed data are migrated to an ESRI geodatabase for spatial analysis and visualization.
- 5 Results: 1,045,785 cardholders had identified housing locations and 362,882 had identified job locations from one-week data.These represent 12.2% and 4.2% of all cardholders, respectively.
- 5 Results: 237,223 cardholders, or 2.8% of the total, had both final housing and job locations.Results were aggregated across bus stops and TAZs, including 3,414 housing-linked and 3,329 job-linked bus stops.
6 Discussion
The discussion presents smart card data as a promising complement to conventional surveys for analyzing Beijing’s commuting patterns, while recognizing important coverage and information limits. Combining LBS and survey data can support urban and transportation planning, but current results remain constrained mainly to bus riders and incomplete bus-route information.
- The study combines rich LBS spatiotemporal information with the social dimensions of conventional surveys to better understand urban dynamics.
- A one-week analysis of individual cardholder bus trips and a decision tree for aggregating daily results produce more solid identification than one-day data.The decision tree uses periodic information and spatial distribution from one-day results.
- The approach retrieves explicit spatial commuting patterns in Beijing with larger samples and more precise spatial and temporal information than conventional questionnaires or household surveys.The authors qualify this comparison by noting that the analysis is limited to bus riders.
- Identified commuting patterns provide information useful for urban and transportation planners, including during multi-year gaps between conventional surveys.
- Limitations: Smart card data exclude trips using other modes, so future studies should combine them with all-modal travel surveys.
- Limitations: Fixed-fare bus routes provide incomplete spatiotemporal information, with some central-area ride information lost and resulting policy implications less accurate.
7 Concluding remarks
The concluding remarks describe a workflow that uses smart card data, survey-derived rules, and urban GIS data to identify jobs, housing, and commuting trips in Beijing. It produces large-scale spatial commuting results, validates them against survey data, and demonstrates feasibility while retaining the scope limits of smart card observations.
- The study proposes TRIP and location-time-duration (PTD) data forms for processing raw smart card data and identifying commuting routes and jobs-housing locations.
- A decision tree combines one-day results into one-week results, providing more accurate identification results.
- 1,045,785 cardholders had identified housing locations, 362,882 had identified job locations, and 237,223 had both.
- 221,773 cardholders had identified commuting trips, which were aggregated by traffic analysis zones using average commuting time and distance.
- The analysis mapped commuting patterns between three residential communities and six business zones and generated dominant links between traffic analysis zones.
- Three-level comparisons with the 2005 household survey produced sound validation results and supported the approach’s applicability.
- The findings demonstrate the feasibility of using smart card data as an alternative to conventional household travel surveys for spatiotemporal urban-structure analysis.