All Issue

2026 Vol.44, Issue 5 Preview Page

Article

A Hierarchical Bayesian Model for Estimating Nationwide Demand for Demand-Responsive Transit at a Fine Spatial Resolution

전국 미시 격자 단위 DRT 이용 수요 추정을 위한 계층적 베이지안 모형

Joonyoung KIM1
,
Gyeonghan KIM1
,
Daecheol KIM1
,
Hyeonsik AN2
,
Tae-Seok KANG3
,
Eui-Jin KIM4*
김 준영1
,
김 경한1
,
김 대철1
,
안 현식2
,
강 태석3
,
김 의진4*
1Student, Department of Transportation Systems Engineering, Ajou University, Gyeonggi 16499, Korea
2Student, Department of Digital Media, Ajou University, Gyeonggi 16499, Korea
3Researcher, InnVation G, Studio Galilei Co. Ltd., Gyeonggi 16827, Korea
4Assistant Professor, Department of Transportation Systems Engineering, Ajou University, Gyeonggi 16499, Korea
1아주대학교 교통시스템공학과 학부과정
2아주대학교 디지털미디어학과 학부과정
3스튜디오 갈릴레이 혁신전략실 책임연구원
4아주대학교 교통시스템공학과 조교수
*Corresponding Author
31 October 2026. pp. 799-819
PDF XML Citation
Abstract

Demand-Responsive Transit (DRT) has been increasingly introduced as a means to complement fixed-route public transport, particularly in rural and small-city areas where conventional transit services are difficult to provide. Estimating DRT demand in potential service areas is essential for selecting priority implementation areas, determining fleet size, and supporting budget planning. However, DRT demand forecasting remains challenging because operational data are limited and demand is sparse and uncertain. This study proposes a hierarchical Bayesian hurdle gamma regression model to estimate the spatial distribution of rural and small-city DRT demand at a nationwide 100m grid resolution. To capture the functional heterogeneity of space, grid cells are clustered into trip-production, trip-attraction, agricultural and industrial, and other functional areas based on building-use composition. The hierarchical structure incorporates cluster-specific effects, while the hurdle structure separately models demand occurrence and positive demand intensity, allowing zero-demand grid cells to be included in the analysis. The results show that population and employment are positively associated with both demand occurrence and positive demand intensity, while the effect of employment varies across building-use clusters. The number of accessible bus stops does not show a statistically clear effect on demand occurrence but has a negative association with positive demand intensity. In terms of predictive performance, the model achieves a Spearman rank correlation of 0.58 at the 100m grid level, with Top 10% and Top 30% hit rates of 41.85% and 62.40%, respectively. When the 100m predictions are aggregated to 500m, 1km, and eup-myeon-dong units, error improves with larger aggregation units. Leave-one-region-out cross-validation further indicates that, although total demand is over- or underestimated in some regions, rank-based performance remains above the random baseline across all test regions. These findings suggest that the proposed model is more suitable for identifying high-potential DRT demand areas and can support policy decisions for rural and small-city DRT implementation and fleet allocation.

Keywords
demand-responsive transit
demand estimation
fine spatial grid
hierarchical bayesian model
rank-based evaluation

수요대응형 대중교통(Demand-Responsive Transit, DRT)은 고정 노선 대중교통 서비스 제공이 어려운 지역을 보완하기 위한 수단으로 확산되고 있다. 신규 도입 지역에 대한 DRT 수요 추정은 사업 지역 선정, 적정 차량 규모 산정, 운영 예산 검토를 위해 중요하지만, DRT는 운영 이력이 축적된 지역이 제한적이고 수요 자체의 불확실성이 높아 미운영 지역의 수요를 예측하는 데 한계가 있다. 이에 본 연구는 제한된 DRT 운영 자료를 활용하여 농어촌 및 중소도시형 DRT 이용 수요의 공간적 분포를 전국 100m 미시 격자 단위에서 추정하기 위한 계층적 베이지안 허들 감마 회귀 모형을 제안한다. 공간의 기능적 이질성을 반영하기 위해 건물의 주용도 면적 비율을 기반으로 격자를 통행생성지, 통행유인지, 농공업지, 기타 지역으로 군집화하고, 군집별 특성을 계층 구조로 반영하였다. 또한 이용량이 0인 격자를 포함하기 위해 DRT 수요 발생 여부와 양의 이용량 규모를 분리하는 허들 구조를 적용하였다. 분석 결과, 인구와 종사자 수는 DRT 수요 발생 여부와 양의 이용량 규모에서 양의 영향을 보였으며, 종사자 수의 영향은 건물 용도 기반 공간 기능에 따라 다르게 나타났다. 접근 가능한 버스 정류장 수는 수요 발생 여부에서는 통계적으로 명확한 효과가 확인되지 않았으나, 양의 이용량 부분에서는 음의 관계를 보였다. 예측 성능 평가 결과, 100m 격자 단위에서 Spearman 순위상관계수는 0.58로 나타났으며, 상위 10% 및 상위 30% 고수요 격자 식별률은 각각 41.85%, 62.40%로 나타났다. 또한 500m, 1km, 읍면동 단위로 예측값을 집계한 결과 오차가 집계 단위가 커질수록 개선되는 경향을 보였다. Leave-one-region-out 교차검증에서도 지역별 총수요 예측에는 과소 또는 과대 추정이 나타났지만, 모든 검증 지역에서 순위 기반 성능은 무작위 기준을 상회하였다. 이러한 결과는 본 모형이 농어촌 및 중소도시형 DRT 도입 후보 지역에서 상대적으로 수요 잠재력이 높은 공간을 식별하기 위한 정책 지원 도구로 활용될 수 있음을 시사한다.

키워드
수요대응형 대중교통(DRT)
수요 추정
미시 격자
계층적 베이지안 모형
순위 기반 평가
References
1

Ahn Y. (2022), Policy Recommendation for DRT Adoption to Urbanized Areas: Based on the ‘Shucle’ Project in Sejong City, Transportation Technology and Policy, 19(3), Korean Society of Transportation, 22-28.

2

Alsaleh N., Farooq B. (2021), Interpretable Data-Driven Demand Modelling for On-Demand Transit Services, Transp. Res. Part A Policy Pract., 154, Elsevier, 1-22.

10.1016/j.tra.2021.10.001
3

Boilé M., Golias M. M., Ivey S. (2010), A Bayesian Hierarchical Network for Truck Demand Modeling, Proceedings of Innovations in Freight Demand Modeling and Data: A Transportation Research Board SHRP 2 Symposium, Washington, D.C., USA.

4

Gelman A., Carlin J. B., Stern H. S., Dunson D. B., Vehtari A., Rubin D. B. (2013), Bayesian Data Analysis,3rd ed.,CRC Press, (Boca Raton, FL, USA), 127.

10.1201/b16018
5

Gelman A., Hill J. (2006), Data Analysis Using Regression and Multilevel/Hierarchical Models, Cambridge University Press, (New York, USA), 251-258.

10.1017/CBO9780511790942
6

Imhof S., Blättler K. (2023), Assessing Spatial Characteristics to Predict DRT Demand in Rural Switzerland, Res. Transp. Econ., 99, Elsevier, 101301.

10.1016/j.retrec.2023.101301
7

Kidando E., Moses R., Sando T., Ozguven E. E. (2019), An Application of Bayesian Multilevel Model to Evaluate Variations in Stochastic and Dynamic Transition of Traffic Conditions, J. Mod. Transp., 27, Springer, 235-249.

10.1007/s40534-019-00199-2
8

Kim E., Oh D., Kim H. (2026), Causal Impacts of Demand Responsive Transit on Public Transit Demand: A Spatial Assessment Framework, Transp. Policy,179 , Elsevier, 104008.

10.1016/j.tranpol.2026.104008
9

Lee J., Choi Y., Kim J. (2024), A Multi-Task Deep Learning Framework for Forecasting Sparse Demand of Demand Responsive Transit, Expert Syst. Appl., 250, Elsevier, 123833.

10.1016/j.eswa.2024.123833
10

Lee J., Kim J., Chung J., Kim J. (2023), The Relationship Between Built Environment and the Ridership of Demand Responsive Transit Using Direct Demand Model: A Case Study of Yeongjong Island, Incheon, J. Korean Soc. Transp., 41(6), Korean Society of Transportation, 690-703.

10.7470/jkst.2023.41.6.690
11

Oh D., Lee E., Kim H., Kim E. (2026), Impact of Bridging Technology on Older Adults’ Use of Demand-Responsive Transit in a Super-Aged Society, Transp. Res. Part A Policy Pract., 204, Elsevier, 104825.

10.1016/j.tra.2025.104825
12

Pak T., Bae B., Lee C., Jung I., Jang B. (2023), Modeling Public Acceptance of Demand-Responsive Transportation: An Integrated UTAUT and ITM Framework, J. Public Transp., 25, Elsevier, 100067.

10.1016/j.jpubtr.2023.100067
13

Park K., Jung H. (2019), User Consciousness Analysis of Bus Alternative DRT, J. Korean Soc. Transp., 37(6), Korean Society of Transportation, 445-457.

10.7470/jkst.2019.37.6.445
14

Pitombeira-Neto A. R., Loureiro C. F. G., Carvalho L. E. (2020), A Dynamic Hierarchical Bayesian Model for the Estimation of Day-to-Day Origin-Destination Flows in Transportation Networks, Netw. Spat. Econ., 20, Springer, 499-527.

10.1007/s11067-019-09490-5
15

Takano H., Morimoto A. (2012), The Difference Between Actual User and Predicted User of Demand Responsive Transit, Journal of Japan Society of Civil Engineers, Ser. D3 (Infrastructure Planning and Management), 68(5), Japan Society of Civil Engineers, I_851-I_856.

10.2208/jscejipm.68.I_851
16

The Seoul Institute (2020), Evaluation of Regional Equity in Public Transportation Services in Seoul (서울시 대중교통 서비스의 지역 형평성 평가), 19.

17

Transportation Research Board (2007), Metropolitan Travel Forecasting: Current Practice and Future Direction, Special Report 288, 86-87.

18

Transportation Research Board (2013), Transit Capacity and Quality of Service Manual, 3rd ed., TCRP Report 165, p. 4-18.

19

Wang H., Li J., Wang P., Teng J., Loo B. P. Y. (2023a), Adaptability Analysis Methods of Demand Responsive Transit: A Review and future directions, Transport Reviews, 43(4), Taylor & Francis, 676-697.

10.1080/01441647.2023.2165574
20

Wang J., Liu K., Yamamoto T., Wang D., Lu G. (2023b), Built Environment as a Precondition for Demand-Responsive Transit (DRT) System Survival: Evidence from an Empirical Study, Travel Behav. Soc., 30, Elsevier, 271-280.

10.1016/j.tbs.2022.10.008
21

Yao X. (2007), Where Are Public Transit Needed - Examining Potential Demand for Public Transit for Commuting Trips, Comput. Environ. Urban Syst., 31(5), Elsevier, 535-550.

10.1016/j.compenvurbsys.2007.08.005
22

Zhang R., Sun Y., Zhang S., Dai Z., Fan J., Han J., et al. (2026), “Clearer” or More “Blurred”? The Evolution of Urban-Rural Boundaries Since the Proposal of Urban-Rural Integrated Development: A Case Study of Zhengzhou, Land, 15(1), MDPI, 195.

10.3390/land15010195
23

Zhang Y., Zheng X., Chen M., Li Y., Yan Y., Wang P. (2021), Urban Fine-Grained Spatial Structure Detection Based on a New Traffic Flow Interaction Analysis Framework, ISPRS Int. J. Geo-Inf., 10(4), MDPI, 227.

10.3390/ijgi10040227
Information
  • Publisher :Korean Society of Transportation
  • Publisher(Ko) :대한교통학회
  • Journal Title :Journal of Korean Society of Transportation
  • Journal Title(Ko) :대한교통학회지
  • Volume : 44
  • No :5
  • Pages :799-819
  • Received Date : 2026-04-08
  • Revised Date : 2026-05-13
  • Accepted Date : 2026-06-22
  • DOI :https://doi.org/10.7470/jkst.2026.44.5.799
Journal Informaiton Journal of Korean Society of Transportation Journal of Korean Society of Transportation
Archives
: 2015 Vol.34~
  • scopus
  • NRF
  • KOFST
  • crossref cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close