ML Models • Rizal

HeatWatch Rizal is an intelligent heat risk monitoring and early warning system powered by Machine Learning and Geographic Information Systems (GIS), designed to protect communities across Rizal Province from extreme heat dangers.

Real-Time Monitoring

HeatWatch continuously tracks temperature and humidity data from all 13 municipalities and 1 component city of Rizal Province, plus 3 nearby Metro Manila reference points, using the Open-Meteo weather API. The interactive GIS map visualizes heat risks in real-time, helping you identify danger zones instantly.

AI-Powered Temperature Forecasting

The system uses three trained machine learning models — Random Forest (95.6% R²), Decision Tree (93.3% R²), and K-Nearest Neighbors (91.9% R²) — to forecast temperature 6 hours ahead. These predictions are then classified into heat risk categories: Normal, Moderate, High, and Dangerous. Switch between models and compare their performance on the ML Models page.

Flexible Alert System

Send heat alerts your way — manually trigger instant email notifications with 6-hour ML forecasts to any recipient, or schedule automatic recurring alerts at your preferred time intervals. Select locations, set your time gap, and let HeatWatch handle the rest.

Data-Driven Insights

All weather readings, risk classifications, ML predictions, and alert logs are stored in a structured database. This continuously growing dataset of real historical weather data serves as a foundation for future model retraining, climate research, and policy-making.

Why It Matters

The Philippines faces increasing heat-related health risks due to climate change. HeatWatch empowers local governments, emergency responders, and communities with actionable data to make informed decisions, issue timely warnings, and ultimately save lives.

Built by Engr. Brian Ezekiel D. Batalon, ECE, ECT, SO2

AI Model Comparison

Live Monitoring

Random Forest

95.6% R² Score

Inactive

Decision Tree

93.3% R² Score

Inactive

K-Nearest Neighbors

91.9% R² Score

Inactive

R² Score Comparison by Model

MAE Comparison (°C) - Lower is Better

RMSE Comparison (°C) - Lower is Better

Predicted vs Actual (Scatter)

Model Error Comparison

Forecasting Model Metrics (6-Hour Prediction)

Model	R² Score	MAE (°C)	RMSE (°C)	Type	Prediction Target
🥇 Random Forest	0.9555	0.58	0.78	Ensemble (100 Trees)	Temperature 6h ahead
🥈 Decision Tree	0.9329	0.73	0.96	Tree-based (Depth 10)	Temperature 6h ahead
🥉 K-Nearest Neighbors	0.9193	0.73	1.05	Distance-based (K=5)	Temperature 6h ahead

Model Information

Random Forest Active

Model	Input Features	Training Data	Data Source	Purpose
Random Forest	Temp, Humidity, Heat Index, Hour, Month, Day	3 Months Historical	Open-Meteo API	Predict temperature 6 hours ahead
Decision Tree	Temp, Humidity, Heat Index, Hour, Month, Day	3 Months Historical	Open-Meteo API	Predict temperature 6 hours ahead
K-Nearest Neighbors	Temp, Humidity, Heat Index, Hour, Month, Day	3 Months Historical	Open-Meteo API	Predict temperature 6 hours ahead