Crowd Management in Delhi Railways: A Data-Driven Approach: Understanding Passenger Flow Patterns for Effective Crowd Management

Abstract

Railway stations in Delhi frequently encounter extreme crowding, particularly during peak festival seasons, posing significant public safety and operational risks. This study is specifically motivated by the tragic crowd crush at New Delhi Railway Station on 15 February 2025, which resulted in 18 fatalities during the Maha Kumbh festival [1]. We analyze passenger footfall and train frequency across five major hubs: New Delhi, Old Delhi, Hazrat Nizamuddin, Anand Vihar Terminal, and Sarai Rohilla. Indian Railways carries approximately 23 million passengers daily [2], making effective density prediction essential. Utilizing a dataset of approximately 10,915 entries, we applied machine learning techniques to predict crowd density. Our analysis identifies critical correlations between train scheduling, seasonal shifts, and dramatic surges in passenger volume during religious observances. Among the models evaluated, the Decision Tree regressor achieved the highest predictive accuracy with a Mean Absolute Error (MAE) of 0.0473, outperforming Random Forest and Linear Regression models. These findings offer a data-driven framework for railway authorities to optimize scheduling and enhance proactive crowd control strategies.