Passenger flow prediction in bus transportation system using deep learning.

Area of Science:

• Transportation Science
• Data Science
• Artificial Intelligence

Background:

• Bus transit systems face challenges in managing passenger flow due to complex operations and unpredictable demand.
• Accurate passenger flow data is essential for optimizing bus scheduling, resource allocation, and improving passenger experience.
• Current methods struggle with the intricacies of passenger movement across various routes and stations.

Purpose of the Study:

• To develop an accurate system for forecasting bus passenger flow within the Karnataka State Road Transport Corporation (KSRTC).
• To enhance operational efficiency and resource planning for the KSRTC Bus Rapid Transit (KSRTCBRT) system.
• To provide reliable predictions for revenue estimation and service optimization.

Main Methods:

• Utilized a deep learning approach combining greedy layer-wise algorithm, Long Short-Term Memory (LSTM), and Recurrent Neural Network (RNN).
• Processed key parameters including bus ID, type, source, destination, passenger count, slot number, and revenue.
• Employed greedy layer-wise algorithm for data clustering into regions, followed by LSTM for data de-duplication and RNN for iterative prediction.

Main Results:

• The proposed deep learning model demonstrated high accuracy in predicting bus passenger flow.
• The system effectively handles the complexities of passenger flow forecasting in a large-scale transit network.
• Achieved improved data processing through clustering and de-duplication techniques.

Conclusions:

• The integrated deep learning framework offers a robust solution for bus passenger flow forecasting.
• The system provides valuable insights for resource planning and revenue estimation in public transportation.
• This approach significantly improves the accuracy and reliability of passenger flow predictions for KSRTC.