Back to Projects
Weather Prediction System
Built a city-specific weather prediction pipeline using XGBoost and historical weather observations, with preprocessing, lag-based feature engineering, and an application interface for accessing predictions.
PythonXGBoostPandasTime SeriesFastAPIDockerGradio
Related Projects
SYSTEM ARCHITECTURE
City-specific XGBoost models are exposed through an API-oriented serving layer.
Problem
City-level weather prediction requires handling temporal patterns and location-specific observations across a large historical dataset.
Architecture
Historical observations → preprocessing and lag features → city-specific XGBoost models → prediction API and user interface.
Challenges
- Prepared historical weather observations for modeling
- Designed lag-based features for time-dependent prediction
- Organized training around city-specific models
- Integrated prediction logic with an API and application interface
Benchmarks
The project uses approximately 10 years of historical data across thousands of cities. Retain dataset-size claims only where they match the documented pipeline.
Lessons Learned
- Feature engineering and temporal validation are important for forecasting
- Location-specific models can capture local patterns
- A prediction interface helps make model outputs accessible