DOI QR코드

DOI QR Code

A study on the number of passengers using the subway stations in Seoul

데이터마이닝 기법을 이용한 서울시 지하철역 승차인원 예측

  • Cho, Soojin (Department of Statistics, Ewha Womans University) ;
  • Kim, Bogyeong (Department of Statistics, Ewha Womans University) ;
  • Kim, Nahyun (Department of Statistics, Ewha Womans University) ;
  • Song, Jongwoo (Department of Statistics, Ewha Womans University)
  • 조수진 (이화여자대학교 통계학과) ;
  • 김보경 (이화여자대학교 통계학과) ;
  • 김나현 (이화여자대학교 통계학과) ;
  • 송종우 (이화여자대학교 통계학과)
  • Received : 2018.09.06
  • Accepted : 2018.11.30
  • Published : 2019.02.28

Abstract

Subways are eco-friendly public transportation that can transport large numbers of passengers safely and quickly. It is necessary to predict the accurate number of passengers in order to increase public interest in subway. This study groups stations on Lines 1 to 9 of the Seoul Metropolitan Subway using clustering analysis. We propose one final prediction model for all stations and three optimal prediction models for each cluster. We found three groups of stations out of 294 total subway stations. The Group 1 area is industrial and commercial, the Group 2 ares is residential and commercial, and the Group 3 area is residential districts. Various data mining techniques were conducted for each group, as well as driving some influential factors on demand prediction. We use our model to predict the number of passengers for 8 new stations which are part of the 3rd extension plan of Seoul metro line 9 opened in October 2018. The estimated average number of passengers per hour is from 241 to 452 and the estimated maximum number of passengers per hour is from 969 to 1515. We believe our analysis can help improve the efficiency of public transportation policy.

지하철은 많은 승객들을 원거리까지 안전하고, 신속 정확하게 원하는 지점으로 대량 수송할 수 있는 친환경적인 교통수단이다. 지하철의 공익성을 증대시키기 위해서는 정확한 승객 수요 예측이 이루어져야 한다. 본 연구는 정확한 지하철 수요예측을 위하여, 군집분석을 통해 서울시 1-9호선 지하철역들을 군집화 하였다. 그 후, 전체 역과 각 군집 별 최종 예측 모형을 제시하였다. 군집화 결과, 294개의 역이 3개로 군집화 되었으며 그룹 1은 상공업지구, 그룹 2는 주상복합지구, 그룹 3은 주거지구가 중심이 되는 역들로 나타났다. 그 후 각 군집 별로 다양한 데이터 마이닝 기법을 이용해 지하철 승차인원 예측 모형을 제시하고, 수요 예측에 중요한 영향을 미치는 요인들을 도출하였다. 그리고 최종 모형을 바탕으로 2018년 10월에 개통될 서울시 9호선 3단계 연장역인 8개 신설역의 3개월 수요를 예측하였다. 8개 신설역의 월평균 시간당 평균 승차인원은 약 241에서 452명, 월평균 시간당 최대 승차인원은 약 969에서 1,515명으로 추정되었다. 본 분석의 최종 모형을 활용한 신설역의 지하철 수요 예측은 대중교통 정책 결정을 위한 기초자료로 활용되어 효율적인 지하철 운영 방안 수립에 기여할 수 있을 것이다.

Keywords

GCGHDE_2019_v32n1_111_f0001.png 이미지

Figure 2.1. Histogram of each response variable.

GCGHDE_2019_v32n1_111_f0002.png 이미지

Figure 2.2. Yearly average of each response variable.

GCGHDE_2019_v32n1_111_f0003.png 이미지

Figure 2.3. Monthly average of each response variable.

GCGHDE_2019_v32n1_111_f0004.png 이미지

Figure 2.4. Response variable vs. Quantile of duration.

GCGHDE_2019_v32n1_111_f0005.png 이미지

Figure 2.5. Response variable vs. Number of subway lines.

GCGHDE_2019_v32n1_111_f0006.png 이미지

Figure 2.6. Response variable vs. Number of exits.

GCGHDE_2019_v32n1_111_f0007.png 이미지

Figure 3.1. Location of subway stations by group.

GCGHDE_2019_v32n1_111_f0008.png 이미지

Figure 3.2. Monthly average of each response variable by group.

GCGHDE_2019_v32n1_111_f0009.png 이미지

Figure 3.3. Yearly average of each response variable by group.

GCGHDE_2019_v32n1_111_f0010.png 이미지

Figure 3.4. Usage area proportion within a radius of 500m around the station.

GCGHDE_2019_v32n1_111_f0011.png 이미지

Figure 3.5. Partial dependence plot of all stations (Model 1).

Table 2.1. Number of subway stations by the number of subway lines

GCGHDE_2019_v32n1_111_t0001.png 이미지

Table 2.2. Description of variables

GCGHDE_2019_v32n1_111_t0002.png 이미지

Table 3.1. Average of predictor by group

GCGHDE_2019_v32n1_111_t0003.png 이미지

Table 3.2. Test root mean squared error of each model using 10-fold cross validation (Model 1)

GCGHDE_2019_v32n1_111_t0004.png 이미지

Table 3.3. Test root mean squared error of each model using 10-fold cross validation (Model 2)

GCGHDE_2019_v32n1_111_t0005.png 이미지

Table 3.4. Test root mean squared error of each group (Model 1)

GCGHDE_2019_v32n1_111_t0006.png 이미지

Table 3.5. Test root mean squared error of each group (Model 2)

GCGHDE_2019_v32n1_111_t0007.png 이미지

Table 3.6. Predicted number of passengers at 8 new stations (2018)

GCGHDE_2019_v32n1_111_t0008.png 이미지

References

  1. Breiman, L., Friedman. J., Olshen, R., and Stone, C. (1984). Classification and Regression Trees, Chapman and Hall, New York.
  2. Breiman, L. (2001). Random forests, Machine Learning, 45, 5-32. https://doi.org/10.1023/A:1010933404324
  3. Cortes, C. and Vapnik, V. (1995). Support-vector networks, Machine Learning, 20, 273-297. https://doi.org/10.1007/BF00994018
  4. Horel, A. E. and Kennard, R. W. (1970). Ridge regression: biased estimation for nonorthogonal problems, Technometrics, 12, 55-67. https://doi.org/10.1080/00401706.1970.10488634
  5. Douglas, R. (2015). Gaussian mixture models, Encyclopedia of biometrics.
  6. Kim, J. I. (2013). The determinants of subway riderships at AM-peak in Daegu metropolitan city: focusing on the land use of station neighborhood areas, Journal of Transport Research, 20, 15-25. https://doi.org/10.34143/JTR.2013.20.1.15
  7. Kim, J. S. (2016). Subway congestion prediction and recommendation system using big data analysis, Journal of Digital Convergence, 14, 289-295. https://doi.org/10.14400/JDC.2016.14.11.289
  8. Lee, J., Go, J. Y., Jeon, S., and Jun, C. (2015). A study of land use characteristics by types of subway station areas in Seoul analyzing patterns of transit ridership, The Korea Spatial Planning Review, 84, 35-53. https://doi.org/10.15793/kspr.2015.84..003
  9. R Development Core Team (2010). R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, ISBN 3-900051-07-0. http://www.R-project.org.
  10. Ridgeway, G. (2007). Generalized Boosted Models: A guide to the gbm package, https://cran.r-project.org/web/packages/gbm
  11. Shon, E. Y., Kwon, B. W., and Lee, M. H. (2004). Modelling the subway demand estimation by station using the multiple regression analysis by category, Journal of Korea Society of Transportation, 22, 33-42.
  12. Song, J. (1991). A study on prediction of passenger demand in Seoul Subway, Statistical Consulting, 6.
  13. Tianqi, C. and Carlos, G. (2016). XGBoost: A Scalable Tree Boosting System, KDD '16 Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785-794.
  14. Tibshirani, R. (1996). Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society B, 58, 267-288.