Passenger Demand Prediction for Metro Station Using Probabilistic Model

Rapid urbanization in India has driven a heightened need for public transportation, particularly Mass Rapid Transit Systems (MRTS) due to their speed, reliability, and affordability. This prompted the initiation of the Delhi Metro project and the preparation of Detailed Project Reports (DPRs) for comprehensive demand assessment. However, a recent audit report revealed significant discrepancies between projected and actual ridership, raising concerns about the existing travel demand modeling method's reliability. Given the high costs of metro system construction and the need for accurate demand estimation, this study focuses on the efficacy of probabilistic models—Gaussian, Negative binomial, and Log-linear regression—for predicting passenger demand at metro stations. The investigation encompasses diverse variables like land use, socio-economic factors, and station attributes. The research concentrates on the Delhi Metro's Yellow Line, a prominent route within India's busiest metro system. To gather the necessary data, the research utilized Geographic Information System (GIS) software to effectively map and analyze the various land use variables. Additionally, a primary survey was conducted to capture relevant passenger characteristics. Once the data collection phase was completed, the research employed statistical software R for the purpose of modeling and analysis. Furthermore, to ensure the robustness of the developed models, additional validation was carried out in two other cities. The study's results indicate that the Gaussian regression model outperforms the negative binomial and log-linear models with an R2 value of 0.93 and a lowest MAPE value of 11%. This suggests that the Gaussian model provides a better fit to the data and more accurately predicts passenger demand. Hence incorporating probabilistic regression models and a comprehensive set of variables can address limitations in the current demand modeling process.