AI-Driven Bus Boarding Demand Prediction for Real-Time Public Transport Optimization

Abstract

Public transportation systems exhibit considerable fluctuations in passenger demand, with over 60% of daily boardings occurring during peak hours and nearly 30% of commuters reporting delays due to overcrowding. In densely populated urban areas, approximately 70% of route optimization efforts fail without accurate boarding demand forecasts. Traditional manual estimation methods—such as timeslot passenger counts, route-based field surveys, and ticket ledger analyses—are often prone to human error, labor-intensive, and lack scalability. These limitations result in inconsistent data, impeding effective transit planning and service optimization. To address these challenges, this study proposes an AI-based bus boarding demand prediction system that leverages Support Vector Machine (SVM) and Deep Neural Network (DNN) classifiers. The models incorporate real-time input features, including timestamps, route numbers, traffic conditions, and boarding volumes. Historical transit datasets are used for training, and model performance is evaluated using standard accuracy metrics to identify the most reliable approach. Furthermore, the predictions are cross-validated against three traditional estimation methods—time-slot monitoring, field-based surveys, and ticket ledger review— to ensure robustness and practical relevance. The resulting framework delivers high-precision, realtime demand predictions, enabling dynamic scheduling and mitigating the mismatch between service availability and commuter demand. This integrated solution bridges the gap in manual estimation approaches and lays the groundwork for a scalable, data-driven public transport optimization system.