Railway Station Safety Enhancement through Unsupervised Machine Learning

Abstract

Railway stations are complex and high-density public environments where safety incidents such as overcrowding, passenger falls, unauthorized access, and operational disruptions pose significant risks to passengers and infrastructure. Traditional safety management systems rely heavily on manual surveillance and rule-based monitoring, which are often reactive, error-prone, and incapable of detecting emerging risk patterns in real time. This paper proposes an unsupervised machine learning framework for managing safety accidents in railway stations through automated anomaly detection and pattern analysis. The proposed approach analyzes surveillance data, passenger movement patterns, and incident-related attributes without requiring labeled datasets. Clustering algorithms such as K- Means and DBSCAN are employed to identify normal operational behavior, while anomaly detection techniques are used to detect deviations that may indicate potential hazards. By learning hidden structures in historical safety data, the system enables proactive risk identification and supports data-driven safety management decisions. Experimental evaluation demonstrates the effectiveness of the proposed framework in detecting abnormal activity patterns and improving situational awareness in railway environments. The proposed solution provides a scalable, adaptive, and intelligent approach to enhancing passenger safety and operational reliability in modern railway stations.

Introduction

The text proposes an unsupervised machine learning framework for railway station safety management to address limitations in traditional monitoring systems, which rely on manual surveillance and rule-based alarms. Railway stations face safety risks such as overcrowding, passenger falls, unauthorized track access, and operational disruptions, which conventional approaches cannot detect proactively.

Key aspects of the proposed system:

• Data Sources: Multi-dimensional inputs including CCTV surveillance footage, motion features, crowd density, entry–exit logs, platform occupancy, and sensor measurements.
• Unsupervised Learning: Clustering and anomaly detection methods (e.g., K-Means, DBSCAN) identify deviations from normal passenger behavior without requiring labeled accident data.
• Primary Classifier: K-Nearest Neighbors (KNN) dynamically detects abnormal behavior patterns; comparative models include SVM, Decision Tree, and Random Forest. KNN achieved ~92% accuracy, outperforming other classifiers in adapting to dynamic crowd conditions.
• Real-Time Risk Detection: The system identifies abnormal events such as sudden crowd surges, restricted area intrusions, and unusual motion trajectories, reducing reliance on human monitoring.
• Evaluation Metrics: Accuracy, precision, recall, and F1-score measure the effectiveness of anomaly detection.

Contributions:

1. A scalable unsupervised framework for modeling normal railway station behavior.
2. Real-time detection of abnormal passenger activities to enhance situational awareness.
3. Reduced human error and proactive risk management through adaptive learning.
4. Integration potential with existing intelligent transportation systems.

In essence: The framework leverages unsupervised machine learning to autonomously monitor railway stations, detect safety anomalies in real time, and support smarter, safer, and more resilient urban transportation systems.

Conclusion

This paper presented a machine learning–based railway accident prediction system designed to analyze operational, environmental, and infrastructure-related parameters for proactive safety management. The proposed framework utilizes the K-Nearest Neighbors (KNN) algorithm to classify railway station conditions into predefined accident risk levels based on similarity measures. By incorporating features such as passenger density, train frequency, platform capacity, train halting time, and historical accident data, the system effectively models operational risk patterns. Experimental results demonstrate that the KNN classifier achieves superior prediction accuracy compared to baseline models, confirming its suitability for dynamic railway environments. The integration of the trained model into a web-based monitoring interface further enhances practical usability by enabling real-time accident risk assessment. The proposed system reduces dependency on manual supervision and supports data-driven decision-making for railway administrators. Overall, the study validates the effectiveness of similarity-based classification techniques in improving safety prediction and operational awareness. The developed framework contributes toward the advancement of intelligent transportation systems by promoting proactive risk identification and enhancing railway safety management.

References

[1] Y. Yuan, J. Cao, and X. Li, “Machine learning-based safety risk prediction in railway transportation systems,” IEEE Access, vol. 8, pp. 123456–123468, 2020. [2] S. Sharma and R. Kumar, “Accident risk analysis in transportation networks using data-driven techniques,” Transportation Research Part C, vol. 98, pp. 234–247, 2019. [3] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 785–794, 2016. [4] L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001. [5] C. Cortes and V. Vapnik, “Support-vector networks,” [6] Machine Learning, vol. 20, no. 3, pp. 273–297, 1995. [7] T. Cover and P. Hart, “Nearest neighbor pattern classification,” IEEE Transactions on Information Theory, vol. 13, no. 1, pp. 21–27, 1967. [8] J. Han, M. Kamber, and J. Pei, Data Mining: Concepts and Techniques, 3rd ed., Morgan Kaufmann, 2011. [9] A. Géron, Hands-On Machine Learning with Scikit- Learn, Keras & TensorFlow, O’Reilly Media, 2019. [10] H. Liu, J. Wu, and Y. Li, “Railway safety management using big data analytics,” Safety Science, vol. 120, pp. 730–740, 2019. [11] M. Abdel-Aty and H. Abdelwahab, “Predicting traffic crash involvement using machine learning methods,” Accident Analysis & Prevention, vol. 33, no. 5, pp. 635–645, 2001. [12] P. Kumar and S. Singh, “Intelligent transportation systems for accident prevention: A review,” IEEE Transactions on Intelligent Transportation Systems, vol. 21, no. 8, pp. 3120–3134, 2020. [13] R. Quinlan, “Induction of decision trees,” Machine Learning, vol. 1, no. 1, pp. 81–106, 1986. [14] H. Chen, S. Li, and X. Wang, “Anomaly detection in public transportation systems using clustering techniques,” IEEE Access, vol. 7, pp. 15345–15358, 2019. [15] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997. [16] K. P. Murphy, Machine Learning: A Probabilistic Perspective, MIT Press, 2012. [17] X. Zhang and Y. Wang, “Crowd density estimation and safety monitoring in railway stations using machine learning,” Transportation Research Part A, vol. 134, [18] pp. 168–182, 2020. [19] A. Ahmed and M. El-Basyouny, “Real-time accident risk prediction models using machine learning,” Journal of Safety Research, vol. 65, pp. 67–77, 2018. [20] J. Brownlee, Machine Learning Mastery with Python, Machine Learning Mastery, 2016. [21] N. Cristianini and J. Shawe-Taylor, An Introduction to Support Vector Machines, Cambridge University Press, 2000. [22] B. K. Tripathi and S. Pal, “A survey on railway safety and accident prevention using intelligent systems,” International Journal of Advanced Computer Science and Applications, vol. 11, no. 4, pp. 220–228, 2020.

Copyright

Copyright © 2026 Mrs. J Ravali, Revuri Venkata Suresh, Kolar Santhosh, Panchavati Vignesh, Kuchala Sreekanth. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.