An AI-driven fire risk forecasting framework for urban villages using IGWO-optimized LSTM with incremental learning

Area of Science:

• Artificial Intelligence
• Urban Planning
• Risk Management

Background:

• Urban villages face significant fire hazards due to high density and informal infrastructure.
• Existing decision-support systems need enhancement for effective urban safety.
• AI offers transformative potential for managing complex urban risks.

Purpose of the Study:

• To develop an AI-driven fire risk forecasting framework for vulnerable urban villages.
• To improve the accuracy and adaptability of fire risk prediction models.
• To integrate advanced optimization and learning strategies for real-time monitoring.

Main Methods:

• An Improved Grey Wolf Optimizer (IGWO) was used to optimize Long Short-Term Memory (LSTM) network hyperparameters.
• An incremental learning (IL) strategy was employed for real-time model updates without full retraining.
• The proposed IGWO-LSTM-IL model was trained and validated using fire incident data from 55 urban villages in Beijing.

Main Results:

• The IGWO-LSTM-IL model achieved a 92.57% reduction in mean squared error compared to a baseline LSTM model.
• The framework demonstrated high predictive accuracy, stability, and adaptability in forecasting fire risks.
• The incremental learning component enabled efficient real-time model updates.

Conclusions:

• The AI-driven IGWO-LSTM-IL framework provides a practical and effective tool for intelligent fire risk monitoring.
• This approach significantly enhances urban safety systems by improving fire hazard prediction.
• The study highlights the potential of AI in transforming urban infrastructure management and risk assessment.