An integrated RISA-machine learning framework to enhance user satisfaction through quality assessment and prioritization of railway facilities

Abstract

The rise in private vehicle usage and resulting negative externalities calls for periodic service quality assessment and improvement of public transport in emerging countries. The present work introduces a novel, integrated RISA-ML (Revised Importance Satisfaction Analysis- Machine Learning) framework to prioritize railway stations and station facilities based on the need for improvement. While station facilities are prioritized using RISA, stations are prioritized based on the Level of Service (LOS) modelled based on ML techniques. The methodology, demonstrated with reference to six railway stations under Indian Railways, combines rigorous data-driven approaches with advanced machine learning techniques. The study identified critical intervention areas, with more than 50% of the facilities in most cases requiring immediate attention. Facilities such as digital information systems and infrastructure for differently abled passengers were emphasized as top priorities. Findings revealed that stations in capital cities, like Trivandrum, exhibited the highest deficiencies, while safety, security, and feeder systems were critical concerns across all stations. A standardized data collection template for LOS model development is presented, ensuring applicability across diverse contexts. Although various Machine learning models viz., Support Vector Regression, Random Forest, and eXtreme Gradient Boost (XGBoost) were applied and rigorously trained, Artificial Neural Network (ANN) emerged as the best fitting LOS model. ANN based feature importance revealed the prominent influence of digital boards, Wi-Fi, and accessibility facilities on LOS. The presented methodology provides actionable insights for systematic infrastructure improvement and offers a scalable solution to enhance passenger satisfaction not only in railway networks but also in other contexts.