Crowd Evacuation in Stadiums Using Fire Alarm Prediction

Area of Science:

• Artificial Intelligence
• Machine Learning
• Public Safety

Background:

• Traditional fire alarms are reactive, leading to delayed responses and increased panic in emergencies.
• High-density environments like stadiums require advanced solutions for efficient and safe evacuations.

Purpose of the Study:

• To develop and evaluate an AI-driven predictive fire alarm and evacuation model for high-occupancy venues.
• To improve emergency response efficiency and public safety by anticipating fire hazards before ignition.

Main Methods:

• Utilized machine learning algorithms with real-time environmental sensor data (62,630 measurements, 15 parameters).
• Compared six models, including Logistic Regression, SVM, Random Forest, and the proposed EvacuNet.
• Focused on early fire risk indicators like temperature, humidity, TVOC, CO2, and particulate matter.

Main Results:

• EvacuNet achieved superior performance with 99.99% accuracy, 1.00 precision, and 1.00 recall.
• The predictive system significantly reduced false alarms and increased fire detection speed.
• AI-driven evacuation optimization minimized congestion and reduced evacuation times.

Conclusions:

• AI-based predictive modeling drastically improves fire response and evacuation efficiency in large-scale venues.
• Intelligent fire detection systems are essential for enhancing public safety in high-occupancy settings.
• Future research should integrate IoT, reinforcement learning, and real-time crowd management for enhanced predictive accuracy.