v0.1.0 — Initial Forecasting MVP

v0.1.0 — Initial Forecasting MVP

Initial release of an end-to-end CPG demand forecasting platform built around a production-style machine learning workflow.

Highlights

• End-to-end forecasting pipeline for retail / CPG demand prediction
• Feature engineering using lag, rolling-window, and calendar-based features
• Baseline model comparison (28-day moving average vs LightGBM)
• Recursive forecasting for future demand prediction
• Model evaluation with MAE, RMSE, and SMAPE
• Interactive Streamlit dashboard for model inspection
• Dockerized setup for reproducible execution
• Automated tests with pytest
• Recruiter-friendly EDA notebook

Included Components

• src/ modular forecasting pipeline
• dashboard/ interactive Streamlit application
• notebooks/ exploratory data analysis notebook
• tests/ automated feature engineering tests
• reports/ generated metrics and visualizations
• Docker support (Dockerfile, docker-compose.yml)
• Project documentation (README.md)

Technical Stack

Python · LightGBM · Streamlit · Docker · Pytest · Pandas · NumPy · Matplotlib · Jupyter

Future Improvements

• External business drivers (holiday, promotion, stockout, marketing features)
• Databricks deployment
• Streamlit cloud deployment
• MLOps-oriented orchestration improvements