Scholarlink

The migration behavior of regular passengers in city transport refers to the movement patterns and preferences of individuals who use public transportation on a regular basis. These passengers may have specific routes, travel times, and preferences for certain modes of transportation. Passenger flow prediction is crucial for understanding and managing this behavior. Accurate predictions enable transportation operators to optimize their services by adjusting vehicle frequency and capacity, deploying additional services during peak periods, providing real-time information, identifying high-demand areas, and expanding the network. This improves the efficiency and reliability of public transportation, catering to the needs of regular passengers. Data science and machine-learning methods allow us to extract correlations from historical data, improving the accuracy of passenger flow prediction. The passenger flow on a station is highly affected by various factors such as the day of the week, holiday, rain, available routes from that station, and some uncertain events like COVID-19. In this study, we have successfully reported passenger flow prediction at various stations using ensemble machine-learning algorithms. Comparative analysis of implemented work has been carried out with the help of statistical parameters and visual infographic details. For accurate prediction model, accurate data cleaning, pre-processing and feature selection based on correlation have been performed on real dataset of Thane Municipal Transport (TMT) from April 2021 to May 2022. We have also compared the performance of prediction models based on month, day and individual station for moderately deviated data and highly deviated data resulted due to effect of COVID-19 pandemic and Taukte cyclone. An exhaustive comparative analysis between train set and test set have been reported with necessary parameters. Comparing to benchmark models, the XGB regressor and Random forest model can reach most accurate prediction and computational efficiency on the real-world dataset. The statistical analysis suggests that all models studied have reported excellent results between train and test data.

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

Station-Wise Boarding Passenger Flow Prediction for Public Transport Using Various Machine-Learning Methods

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Madhuri Patel,
  • Samir B. Patel,
  • Debabrata Swain,
  • Shubh Patel

摘要

The migration behavior of regular passengers in city transport refers to the movement patterns and preferences of individuals who use public transportation on a regular basis. These passengers may have specific routes, travel times, and preferences for certain modes of transportation. Passenger flow prediction is crucial for understanding and managing this behavior. Accurate predictions enable transportation operators to optimize their services by adjusting vehicle frequency and capacity, deploying additional services during peak periods, providing real-time information, identifying high-demand areas, and expanding the network. This improves the efficiency and reliability of public transportation, catering to the needs of regular passengers. Data science and machine-learning methods allow us to extract correlations from historical data, improving the accuracy of passenger flow prediction. The passenger flow on a station is highly affected by various factors such as the day of the week, holiday, rain, available routes from that station, and some uncertain events like COVID-19. In this study, we have successfully reported passenger flow prediction at various stations using ensemble machine-learning algorithms. Comparative analysis of implemented work has been carried out with the help of statistical parameters and visual infographic details. For accurate prediction model, accurate data cleaning, pre-processing and feature selection based on correlation have been performed on real dataset of Thane Municipal Transport (TMT) from April 2021 to May 2022. We have also compared the performance of prediction models based on month, day and individual station for moderately deviated data and highly deviated data resulted due to effect of COVID-19 pandemic and Taukte cyclone. An exhaustive comparative analysis between train set and test set have been reported with necessary parameters. Comparing to benchmark models, the XGB regressor and Random forest model can reach most accurate prediction and computational efficiency on the real-world dataset. The statistical analysis suggests that all models studied have reported excellent results between train and test data.