Geospatial Predictive Analytics for Enhancing Fire Response in Sta. Cruz Laguna

Abstract

Fire response in urbanizing areas like Sta. Cruz, Laguna, in particular, wrestles with challenges like delayed response times, suboptimal routing, and limited access to fire hydrant points which have all been largely caused by manual dispatching and static navigation. This article reports the title “Geospatial Predictive Analytics for Enhancing Fire Response in Sta. Cruz, Laguna," a web-based intelligent system that integrates machine learning with Geographic Information Systems (GIS) to optimize fire emergency operations. Using a developmental and experimental research approach, the investigators trained an ensemble of models - XGBoost, Random Forest, and Gradient Boosting - on 2,180 historical fire incident records, incorporating engineered features like casualty indices, false alarm indicators, and environmental factors (temperature, precipitation, road conditions). The system runs on a Flask backend with SQLAlchemy for data persistence, housing geospatial datasets such as fire hydrant coordinates and hazardous road segments.
The ensemble model delivered a mean absolute error (MAE) of 0.573 minutes (about 34 seconds) and an R² score of 0.877, surpassing the target accuracy threshold. Its integrated GIS component offers dynamic visualization of optimized routes and nearest hydrant recommendations.
The interactive dashboard provides real-time estimated time of arrival (ETA) predictions, automated performance comparisons, and actionable feedback to drive operational improvements. In the end, this innovation shifts manual, intuition-driven dispatching to a data-backed decision-support system, delivering measurable gains in response efficiency for fire protection agencies.