Disaster Response AI: Resilience Net Forecaster – Predictive Hazard Mapping & Resource Allocation

Disaster management systems worldwide have historically operated reactively, mobilizing resources only after catastrophic events have already unfolded. ResilienceNet Forecaster addresses this critical gap by presenting an AI-driven Disaster Risk Management (DRM) platform that integrates Machine Learning (ML), Explainable Artificial Intelligence (XAI), geospatial hazard mapping, and intelligent resource allocation into a unified, deployable system. The proposed system analyzes the FEMA Disaster Declarations Summaries dataset augmented with environmental and demographic attributes, applying a robust preprocessing pipeline, Exploratory Data Analysis (EDA), and Random Forest algorithms (Classifier and Regressor) to predict disaster risk levels and estimate potential impact. Feature importance plots with reversible label encoding provide transparent, interpretable explanations aligned with XAI-DRM research consensus. Interactive choropleth and point-based hazard maps render geospatial risk distributions, while a proportional allocation engine converts risk scores into actionable emergency response plans covering rescue teams, supply kits, and evacuation capacity. The system, implemented in Python using Streamlit, Scikit-learn, Pandas, Plotly, and Matplotlib, achieves a classification accuracy of 89% and an F1-score of 0.87 on historical disaster data. All modules are integrated within a single web interface supporting end-to-end workflows from data ingestion to downloadable CSV resource plans. This work directly operationalizes the theoretical XAI-DRM roadmap of Ghaffarian et al. (2023) into a practical open-source platform suitable for FEMA offices, state Emergency Operations Centers (EOCs), NGOs, and humanitarian organizations.