Age-sensitive urban rail passenger demand forecasting and uncertainty-driven anomaly detection using a hybrid...

Area of Science:

• Urban planning and transportation science
• Data science and machine learning
• Predictive modeling and anomaly detection

Background:

• Urban demographics are rapidly evolving, necessitating advanced predictive modeling for rail transit systems.
• Optimizing rail transit capacity and reliability is crucial for efficient urban mobility.
• Existing forecasting models may not adequately capture age-specific travel behaviors.

Purpose of the Study:

• To develop and evaluate a novel age-sensitive demand forecasting and anomaly detection framework for Istanbul's urban rail network.
• To benchmark the performance of a proposed hybrid ensemble model against established machine learning algorithms.
• To provide actionable insights for intelligent transportation planning and real-time capacity management.

Main Methods:

• Utilized a dataset of 721,328 passenger-trip records (2021-2023) from Istanbul's urban rail network.
• Engineered eleven spatiotemporal and transactional features to classify passengers into four age cohorts.
• Developed a two-stage hybrid ensemble integrating the SAINT Transformer and CatBoost, enhanced with calibrated uncertainty meta-features.

Main Results:

• The proposed hybrid ensemble achieved a new state-of-the-art accuracy of 91.94% and ROC-AUC of 0.9910.
• Significantly outperformed standalone SAINT (90.12%) and CatBoost (74.78%) models, with statistical significance confirmed by McNemar's test (p < 0.001).
• Introduced an unsupervised anomaly detection mechanism achieving a ROC-AUC of 0.77 and provided interpretable insights via SHAP analysis.

Conclusions:

• The developed framework offers a robust, calibrated, and interpretable solution for intelligent urban rail transit planning.
• Age-sensitive demand forecasting and anomaly detection are critical for optimizing capacity and enhancing operational resilience.
• The findings provide actionable insights for real-time management of urban transportation networks.