Shams-Saied/weather-predictor-django

Django web app that serves a SARIMAX + Fourier-seasonality temperature forecast for Alexandria, Egypt — pick a date, get a prediction with a 95% CI. # Alexandria Weather Forecast — Django App (Task 6) A Django web app that wraps the SARIMAX temperature-forecasting model from **Task 5** (`weather_forecasting_sarima.ipynb`) so a user can pick a date and instantly get a predicted daily mean temperature for Alexandria, Egypt, with a 95% confidence interval. | | | |---|---| | Form | | | Result | | | Validation error | | ## What's here ``` weather_predictor/ ├── manage.py ├── requirements.txt ├── weather_project/ # Django project settings/urls ├── predictor/ # The forecasting app │ ├── forms.py # Date input + validation │ ├── views.py # Form view + JSON API + caching │ ├── model_loader.py # Loads the pickled SARIMAX model once, builds │ │ # future Fourier exog, runs get_forecast() │ ├── ml/train_model.py # Reproduces the Task 5 pipeline end-to-end │ │ # and saves model/sarimax_model.pkl │ ├── templates/predictor/ # Bootstrap 5 templates │ └── static/predictor/css/ # Custom styling on top of Bootstrap ├── model/ │ ├── sarimax_model.pkl # Trained model (statsmodels SARIMAXResults) │ └── model_metadata.json # order, Fourier settings, training date range ├── data/weather_raw.json # Offline data cache (see note below) └── screenshots/ ``` ## About the bundled model `data/weather_raw.json` is the real NASA POWER pull from Task 5 — daily temperature, precipitation, humidity, and wind speed for Alexandria, Egypt, 2021-01-01 through 2025-12-31. `model/sarimax_model.pkl` was trained directly on this data by `predictor/ml/train_model.py`, refitting the notebook's `SARIMAX(1,1,2)` + annual Fourier terms spec on the full history so the app's forecasts start from the most recent real observation (2025-12-31). `predictor/ml/train_model.py` tries the live NASA POWER API first and only falls back to this cached file if the request fails (e.g. no network access), exactly like Task 5's …