An Intelligent Model for Predicting and Preventing Overcrowding Incidents in Holy Sites Using Artificial Intelligence Technologies

Abstract

Background: The Hajj pilgrimage, conducted annually, is one of the largest religious gatherings in the world, with over 2 million pilgrims converging on the sacred sites in Saudi Arabia. Incidents registered during the 1994-2015 period have also demonstrated that the reality of predictive Challenge management systems is indeed essential in eliminating incidents resulting from crowd behavior.
Objective: This study develops and validates an AI-powered Challenge prediction model designed explicitly for Hajj pilgrimage management, utilizing deep learning techniques to forecast potential crowd emergencies and enable proactive intervention strategies.
Techniques: A comprehensive 30-year dataset (1994-2024) was developed, based on historical incident data, real-time surveillance feeds, environmental sensors, and pilgrim flow patterns. A hybrid architecture combining Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN) with ensemble techniques enables the integration of temporal analysis and spatial pattern identification within a single structure, resulting in robust classification.
Results: The model proposed in this paper attained an overall accuracy of 87.3% in predicting crises, with 91.2% accuracy in predicting major incidents. The system shows a false positive rate of 8.1 percent and a false negative rate of 4.7 percent, with an average lead prediction time of 2.3 minutes. The evaluation of performance using the Area Under the Curve (AUC-ROC) yielded a value of 0.89, indicating excellent discriminative ability.
Conclusion: The current research is the first to provide a comprehensive artificial intelligence-based Challenge prediction system for religious mass events. The model's accuracy, fast response, and cultural sensitivity enable it to foster safety in sacred spaces.