Iowa Liquor Sales - Forecasting Time Series Retail Data using XGBoost

How effectively can a trained model predict retail sales of liquor for a store in the state of Iowa?

Data used for this project:

Iowa Liquor Sales, available as a public dataset on Google's BigQuery, or here: https://data.iowa.gov/Sales-Distribution/Iowa-Liquor-Sales/m3tr-qhgy

EDA Questions explored:

1. Which store or chain of stores sold the most alcohol (by total dollar amount of sales) in 2019 and 2020?
2. What are the highest purchased items, and highest purchased categories (by total dollar amount of slaes) at Hy-Vee locations in Iowa for 2020?
3. For the Hy-Vee location in Cedar Falls, which month, days of the month, and days of the week had the highest sales totals (measured in dollars) for 2020?

Model Questions explored:

1. Once add'l date features, daily, weekly, and monthly averages are added, which features correlate most strongly with sales amount in dollars?
2. When an XGBoost model is trained on the 2020 data for the Hy-Vee in Cedar Falls, how accurate can it forecast the 2021 sales data?

Please note: originally created in a notebook instance on Google Cloud Platform