All Issue

2025 Vol.43, Issue 2 Preview Page

Article

Taxi Demand Forecasting in Seoul Considering the Nonlinear Characteristics of Explanatory Variables: Application of the Multivariate Fractional Polynomial Model

설명변수의 비선형적 특성을 고려한 서울시 택시이용수요예측 분석: Multivariate Fractional Polynomial Model의 적용

Minsu YANG1
,
Jungyeol HONG2*
양 민수1
,
홍 정열2*
1Master Course, Department of Transportation Engineering, Keimyung University, Daegu 42601, Korea
2Assistant Professor, Department of Transportation Engineering, Keimyung University, Daegu 42601, Korea
1계명대학교 교통공학과 석사과정
2계명대학교 교통공학과 조교수
*Corresponding Author
30 April 2025. pp. 197-209
PDF XML Citation
Abstract

Various factors have intricately influenced taxi demand, including social, environmental, and economic conditions, policy changes, and unpredictable events. Notably, policies such as social distancing have caused dramatic fluctuations in taxi demand since the onset of the COVID-19 pandemic in 2020. Recently, the activation of mobility platforms has intensified competition in the taxi market. Previous studies have predominantly focused on taxi demand forecasting based on historical time-series trends. However, such approaches often need to account for rapid social changes and event-driven factors adequately. This study develops a daily taxi demand forecasting model for Seoul by applying the Multivariate Fractional Polynomial (MFP) method to address these limitations. This approach effectively captures the complex, nonlinear relationships between variables through variable transformation, providing a refined explanation of the underlying data dynamics. The analysis identified significant factors affecting daily taxi demand, including residential population, floating population, temperature, precipitation, public transportation usage (bus and subway boardings and alightings), and confirmed COVID-19 cases. The findings indicate that taxi demand increases with higher precipitation and temperature levels and more COVID-19 cases. The MFP model demonstrated superior predictive performance compared to traditional linear regression models in key performance metrics, substantiating its reliability by accurately reflecting the nonlinear characteristics of the data. The taxi demand forecasting model developed in this study is expected to provide practical support for policymakers and industry practitioners. Specifically, it can serve as a data-driven tool to address imbalances between taxi supply and demand, support efficient dispatch strategies, and aid in the design of demand-oriented policies.

Keywords
demand forecasting model
multivariate fractional polynomial
nonlinear relationships
social, environmental, and economic factors
taxi demand
variable transformation

택시 이용수요는 사회·환경·경제적 요인, 정책 변화, 예측 불가능한 이벤트 등 다양한 요소에 의해 복잡하게 변화해 왔다. 특히, 2020년 COVID-19 팬데믹 이후 사회적 거리두기 등의 정책이 시행되며 택시 이용수요가 급격히 변화하였으며, 최근 모빌리티 플랫폼의 활성화는 택시 시장의 경쟁을 심화시키고 있다. 기존 연구들은 과거의 시계열적 추세를 기반으로 하거나 데이터 학습을 통한 택시 이용 수요예측 분석을 주로 수행해왔으나, 이러한 접근 방식은 급변하는 사회적 변화와 예측이 불가능한 이벤트 요인들을 충분히 반영하지 못하는 한계를 지닌다. 이에 본 연구는 택시 이용수요의 영향을 주는 변수 간의 비선형적 관계를 정교하게 반영할 수 있도록, 변수 변환을 통해 데이터에 내재된 다양한 관계를 효과적으로 설명할 수 있는 다변량 분수 다항식(Multivariate Fractional Polynomial, MFP)을 적용하여 서울시 일일 택시 이용 수요예측 모델을 제시하였다. 분석 결과 생활인구, 유동인구, 기온, 강수량, 대중교통 이용량(버스 및 지하철 승·하차 건수), COVID-19 확진자 등이 일일 택시 이용 수요에 유의한 영향을 미치는 변수로 도출되었으며, 특히 강수량 및 기온이 높고, 코로나 확진자 수가 많을수록 택시의 이용수요는 높아지는 경향을 나타내었다. 다변량 분수 다항식 모델은 예측 성능 지표에서 선형 회귀모델 대비 우수한 성능을 보여주었으며, 데이터의 비선형적 특성을 정교히 반영함으로써 모델의 신뢰성을 입증하였다. 본 연구에서 도출된 택시 이용 수요예측 모델은 정책 입안자와 운수업계 실무자들에게 실질적인 도움을 제공할 수 있다. 특히, 택시 공급과 수요 간의 불균형을 해소하고, 효율적인 배차 전략 및 수요 기반 정책 설계를 지원할 수 있는 데이터 기반 도구로 활용될 수 있을 것이다.

키워드
수요예측모델
다변량 분수 다항식
비선형적 관계
사회·환경·경제적 영향요인
택시 수요
변수 변환
References
1

Askari B., Le Quy T., Ntoutsi E. (2020), Taxi Demand Prediction Using an LSTM-Based Deep Sequence Model and Points of Interest, In 2020 IEEE 44th Annual Computers, Software, and Applications Conference (COMPSAC), 1719-1724, IEEE.

10.1109/COMPSAC48688.2020.000-7
2

Box G. E., Tidwell P. W. (1962), Transformation of the Independent Variables, Technometrics, 4(4), 531-550.

10.1080/00401706.1962.10490038
3

Gangrade A., Pratyush P., Hajela G. (2022), Taxi‐Demand Forecasting Using Dynamic Spatiotemporal Analysis, ETRI Journal, 44(4), 624-640.

10.4218/etrij.2021-0123
4

Hauer E. (2015), The Art of Regression Modeling in Road Safety, 38, New York, Springer.

10.1007/978-3-319-12529-9
5

Kim S. S., Kang N. W. (2019), A Preliminary Study on the Digital Twin Design for Autonomous Taxi Services in Seoul: Passenger Demand Forecasting Using Deep Learning and Big Data, Proceedings of the Korean Society of Mechanical Engineers Spring and Autumn Conference, 2273-2274.

6

Kim Y. W., Kim Y. J., Kim J. A., Lee Y. J., Lee W. G., Lee S. A. (2021), Demand Forecasting and Improvement of Taxi Stand Locations Through Taxi Data Analysis, Proceedings of the Korean Institute of Information Scientists and Engineers Conference, 1264-1266.

7

Lee E. K., Hwang H. J., Choi H. S., Kim D. K. (2022), A Study on Demand Forecasting Methods for Disabled Call Taxis in Seoul by Administrative Districts Using Deep Learning, The 86th Conference of KST, Korean Society of Transportation, 316-317.

8

Liu Z., Chen H. (2022), Short-Term Online Taxi-Hailing Demand Prediction Based on the Multimode Traffic Data in Metro Station Areas, Journal of Transportation Engineering, Part A: Systems, 148(6), 05022003.

10.1061/JTEPBS.0000681
9

Luo A., Shangguan B., Yang C., Gao F., Fang Z., Yu D. (2022), Spatial-Temporal Diffusion Convolutional Network: A Novel Framework for Taxi Demand Forecasting, ISPRS International Journal of Geo-Information, 11(3), 193.

10.3390/ijgi11030193
10

Royston P., Altman D. G. (1994), Regression Using Fractional Polynomials of Continuous Covariates: Parsimonious Parametric Modelling, Journal of the Royal Statistical Society Series C: Applied Statistics, 43(3), 429-453.

10.2307/2986270
11

Royston P., Sauerbrei W. (2003), Stability of Multivariable Fractional Polynomial Models With Selection of Variables and Transformations: A Bootstrap Investigation, Statistics in Medicine, 22(4), 639-659.

10.1002/sim.131012590419
12

Yoon Y. B., Kim N. Y., Ryu M. J., Yang H. J., Hong Y. R. (2018), A System for Providing Information on Taxi Passenger Demand Forecasting and Movement Pattern Analysis Using Big Data, Proceedings of the Korean Institute of Industrial Engineers Fall Conference, 2822-2836.

13

Zhao J., Kong W., Zhou M., Zhou T., Xu Y., Li M. (2022), Prediction of Urban Taxi Travel Demand by Using Hybrid Dynamic Graph Convolutional Network Model, Sensors, 22(16), 5982.

10.3390/s2216598236015740PMC9415392
Information
  • Publisher :Korean Society of Transportation
  • Publisher(Ko) :대한교통학회
  • Journal Title :Journal of Korean Society of Transportation
  • Journal Title(Ko) :대한교통학회지
  • Volume : 43
  • No :2
  • Pages :197-209
  • Received Date : 2024-12-02
  • Revised Date : 2025-01-03
  • Accepted Date : 2025-02-12
  • DOI :https://doi.org/10.7470/jkst.2025.43.2.197
Journal Informaiton Journal of Korean Society of Transportation Journal of Korean Society of Transportation
Archives
: 2015 Vol.34~
  • NRF
  • KOFST
  • crossref cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close