onyekaugochukwu/sktime-pycon2025-presentation

This repo contains code used in sktime PyCon Nigeria 2025 demo # 🥐 Retail Forecaster This project is an interactive Streamlit app for forecasting daily croissant sales using real historical data and state-of-the-art time series models from the `sktime` library. The app provides three levels of forecasting sophistication, from manual model selection to automated hyperparameter optimization. ## Features - **Level 1: The Craftsman** – Manually select and run a forecasting model. - **Level 2: The Foreman** – Automatically compare several models and pick the best. - **Level 3: The Optimizer** – Fine-tune model hyperparameters for optimal accuracy. - Visualizes historical sales and forecast results interactively. ## Project Structure - `app.py` – Main Streamlit application. - `croissant.csv` – Daily croissant sales data (with `date` and `croissant_sales` columns). - `requirements.txt` – Python dependencies. ## How to Run 1. **Install dependencies** (preferably in a virtual environment): ```bash pip install -r requirements.txt ``` 2. **Ensure the data file** `croissant.csv` is in the same directory as `app.py`. 3. **Start the Streamlit app:** ```bash streamlit run app.py ``` 4. **Open the app** - Streamlit will provide a local URL (e.g., `localhost`) in your terminal. Open it in your browser. ## Requirements - Python 3.8+ - See `requirements.txt` for all required packages. ## Notes - If you add new data, ensure it follows the same format as `croissant.csv`. - The app uses caching for fast data loading. - For best results, use the latest version of `sktime` and `streamlit`. --- **Enjoy forecasting and growing your croissant business! 🥐**