Scholarlink

<p>Inaccurate bus passenger demand prediction hinders the efficient allocation of resources within urban transportation systems, consequently leading to passenger dissatisfaction. While prior studies have explored demand forecasting, few have comprehensively integrated the influence of environmental factors and dynamically categorized daily patterns. To address the challenge of capturing these multifaceted demand drivers, this study proposes a novel machine learning framework that holistically incorporates Automatic Fare Collection (AFC) data, meteorological variables, the Air Quality Index (AQI), and day type classifications derived through K-Means clustering. Five distinct models ANFIS, RFR, GBM, SVR, and KNN were rigorously evaluated using a comprehensive dataset spanning one year and encompassing 155 bus lines in Mashhad, Iran. The Random Forest Regression (RFR) model demonstrated the most robust performance, achieving a coefficient of determination (R2) of 0.739 and a Mean Absolute Percentage Error (MAPE) of 27.9%, thereby outperforming the alternative methods. Subsequent validation on a separate set of 64 bus lines confirmed the framework’s real-world applicability, yielding an average Root Mean Square Error (RMSE) below 10 passengers per hour. These findings underscore the significant value of integrating environmental and event-driven variables into demand forecasting models, offering a robust tool for optimizing bus scheduling and ultimately enhancing the passenger experience within public transportation systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Machine Learning Framework for Daily Bus Passenger Demand Prediction Considering Weather, Air Quality, and Day Type Effects

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Saeed Rashvand,
  • Ali Gholami

摘要

Inaccurate bus passenger demand prediction hinders the efficient allocation of resources within urban transportation systems, consequently leading to passenger dissatisfaction. While prior studies have explored demand forecasting, few have comprehensively integrated the influence of environmental factors and dynamically categorized daily patterns. To address the challenge of capturing these multifaceted demand drivers, this study proposes a novel machine learning framework that holistically incorporates Automatic Fare Collection (AFC) data, meteorological variables, the Air Quality Index (AQI), and day type classifications derived through K-Means clustering. Five distinct models ANFIS, RFR, GBM, SVR, and KNN were rigorously evaluated using a comprehensive dataset spanning one year and encompassing 155 bus lines in Mashhad, Iran. The Random Forest Regression (RFR) model demonstrated the most robust performance, achieving a coefficient of determination (R2) of 0.739 and a Mean Absolute Percentage Error (MAPE) of 27.9%, thereby outperforming the alternative methods. Subsequent validation on a separate set of 64 bus lines confirmed the framework’s real-world applicability, yielding an average Root Mean Square Error (RMSE) below 10 passengers per hour. These findings underscore the significant value of integrating environmental and event-driven variables into demand forecasting models, offering a robust tool for optimizing bus scheduling and ultimately enhancing the passenger experience within public transportation systems.