Project Resources
These are the libraries, data sources, and tools that power our APAW project.
Core Algorithms & Libraries
Machine learning algorithm used for classification of flood/no flood in training and prediction.
Long Short-Term Memory networks used for time-series regression predicting flood height in training and inference.
Provides foundational ML algorithms, model tuning, and evaluation tools used during training.
Deep learning framework used for building and training the LSTM flood height regression model.
Used extensively for time-series and tabular data manipulation in training data preparation and inference.
Fundamental numerical computing library utilized for array operations during model training and inference.
Employed for geospatial data processing and offline preparation of geographic features used as model inputs.
Used for manipulation and analysis of planar geometric objects, instrumental in geospatial feature engineering.
Used for efficient serialization and loading of trained machine learning models and preprocessing pipelines.
Python library used to fetch external API data such as weather and live water levels during inference.
Used for accurate timezone conversions in handling date-time data during preprocessing and inference.
Ancillary AI Models
Referenced as an AI model for parsing external PAGASA cyclone data, not integrated in backend model training or prediction.
Data Sources
Primary source for real-time and historical water level station data, tropical cyclone data, and flood advisories.
Provided elevation data used to calculate geospatial features like elevation difference to waterways.
Supplementary weather forecasting data source integrated into model features and user interface displays.
Provider of geographic baseline features like waterways, used in offline spatial feature engineering and context extraction.
Source for SRTM30m elevation data, used to derive key geospatial features.
Historical Flood Event Records
Aggregated from government advisories and public records, used for training label generation and evaluation.
Development & Deployment Stack
Cloud-based development environment utilized for initial data processing, model training, and model evaluation experiments.
Framework used for rapid prototyping and deployment of the frontend prediction interface.
Backend service platform storing pre-fetched weather and soil forecast data needed for model input features.
Framework for building the frontend application that visualizes flood predictions.
JavaScript library enabling interactive maps within the frontend user interface.
OpenStreetMap-based geocoding service utilized for location search features in the frontend.
OpenStreetMap-based geocoding API for querying coordinates for search functionality.
Other Tools
Web application utilized during data collection for organizing location data.
GitHub repository providing administrative boundaries in GeoJSON for visualization and context.
Institution
Academic institution providing project support for the capstone.