About Us
A machine learning-based predictive analytics platform for flood risk assessment, applying trained models to evaluate complex hydrometeorological data for Metro Manila.
What is 
?
APAW (Advanced Predictive Analysis of Water-related Flood Risk) is a machine learning-based predictive analytics platform developed as a capstone project. By leveraging pattern-matching algorithms and learning models, APAW interprets weather datasets, geographical features, and historical timelines to estimate potential flood risks.
Designed to complement official government systems, our experimental infrastructure explores localized insights for specific coordinates across Metro Manila, aiming to aid in community awareness and modern disaster preparedness education.
System Benchmarks
Our machine learning models were evaluated on training data to establish baseline performance metrics for our capstone research.
Training Classification
During initial model training, our flood detection algorithm achieved 88% accuracy in determining potential flood conditions, alongside a 73% recall rate.
Depth Estimation Metrics
When testing depth predictions, the LSTM model recorded a Mean Absolute Error of 8.07 cm on our historical data sets, providing a baseline for localized estimations.
Hourly Forecast Window
We aim to provide hourly predictions up to 5 days in advance, a research approach intended to expand preparation windows.
Fast Response Time
Our system is computationally optimized to deliver predictive calculations within 5-15 seconds for timely data analysis.
The Methodology
A four-step pipeline combining data aggregation with predictive machine learning models.
Data Integration
We combine weather forecasts, historical flood records, and geographic features from multiple reliable sources to create a comprehensive dataset.
Data Integration
We combine weather forecasts, historical flood records, and geographic features from multiple reliable sources to create a comprehensive dataset.
Flood Detection with Random Forest
Our Random Forest model (a pattern-matching algorithm) analyzes patterns in the integrated data to estimate the likelihood of flooding at a specific location and time.
Flood Detection with Random Forest
Our Random Forest model (a pattern-matching algorithm) analyzes patterns in the integrated data to estimate the likelihood of flooding at a specific location and time.
Depth Prediction with LSTM
When a potential flood is detected, our LSTM model (which learns from historical sequences) processes temporal data to estimate flood depth based on past water level patterns.
Depth Prediction with LSTM
When a potential flood is detected, our LSTM model (which learns from historical sequences) processes temporal data to estimate flood depth based on past water level patterns.
Real-time Aggregation
The system continuously processes recent weather conditions to generate updated hourly flood risk assessments up to 5 days ahead for locations across Metro Manila.
Real-time Aggregation
The system continuously processes recent weather conditions to generate updated hourly flood risk assessments up to 5 days ahead for locations across Metro Manila.
A Localized Approach to Preparedness
How APAW explores targeted, location-specific flood risk assessment.
Traditional Systems
- Prediction Timeframe Same-day or short-term warnings (hours ahead)
- Coverage Area Regional coverage (e.g., NCR flood risk)
- Information Provided General flood alerts
- Accessibility Government announcements and media
APAW Architecture
- Prediction Timeframe Hourly forecasts up to 5 days in advance
- Coverage Area Any specific location in Metro Manila
- Information Provided Specific flood depth predictions
- Accessibility Public web platform
Our Core Philosophy
The principles stitched into the very acronym of our identity.
Accuracy
We are committed to thorough analysis and continuous improvement to provide reliable flood risk assessments, striving to make our experimental models as helpful as possible for academic research.
Preparedness
We aim to empower individuals and communities with accessible information that fosters awareness and promotes proactive planning for potential localized flood events.
Accessibility
We strive to make complex flood risk data understandable and readily available to everyone through user-friendly interfaces and clear, community-focused communication.
Wisdom
We leverage data-driven insights and machine learning concepts to help translate raw statistical information into accessible knowledge for communities.
The Mission
To contribute to community resilience research through accessible, predictive data models. We visualize a platform that can eventually help individuals and communities explore potential localized flood scenarios ahead of time.
The Vision
To explore how localized machine learning can mitigate the impact of unexpected weather events, fostering a future where communities are better informed about data-driven risk assessment within Metro Manila.