trydoff/Product-Demand-Forecasting-Using-ML — reverse-engineered prompt

Build me a simple Python notebook that uses the included sales data to forecast product demand, basically predicting units sold from past records like week, store, SKU, price, and whether an item was featured or on display.

I want this to feel practical for a supply chain use case, not just a toy example. Please clean up missing values, pick sensible features, turn the historical sales data into a format that works for machine learning, and train a couple of forecasting models such as Random Forest and XGBoost. If it makes sense, do some basic tuning and compare the models clearly so I can see which one works better.

Show the results with easy charts and error metrics, and include a short explanation of what the model is learning and how someone could use it for planning inventory. If anything in the data needs assumptions, make reasonable choices and explain them in comments. You can look up current docs online if needed.