Asteroid Collision Risk Prediction
A Random Forest architecture designed to predict asteroid collision risks using live telemetry from NASA APIs.
Near-Earth Object Surveillance
Developed in 2025, this project implements a Random Forest (RF) machine learning model to evaluate and predict the risk of collision with Near-Earth Objects (NEOs).
Data Engineering & NASA APIs
The system bypasses static datasets by interfacing directly with the NASA JPL NeoWs (Near Earth Object Web Service) APIs. It pulls live orbital data, structuring a robust telemetry pipeline. Features engineered for the model include:
- Absolute magnitude (luminosity-based size estimation).
- Relative velocity (km/s).
- Miss distance (astronomical units and lunar distances).
- Orbit uncertainty parameters.
Modeling Approach
I chose a Random Forest architecture for its inherent ability to handle non-linear planetary data and its robustness against overfitting. By analyzing hundreds of thousands of historical NEO trajectories, the model establishes classification boundaries between hazardous and non-hazardous objects.
The project demonstrates end-to-end data science capabilities: from live API data ingestion and cleaning to feature engineering, model training, and inference.