Iowa Liquor Sales Forecast

This repository contains the functions created to generate a sales forecasting model that predicts sales based on the historical data of liquor purchases from the state of Iowa.

The created model consists of a multivariate ARIMA model that includes relevant features such as moving averages of key columns from the dataset, lag columns and weather forecast information.

All data used to train the model was obtained from the library of BigQuery public datasets.

All the datasets and models created are stored inside BigQuery. Therefore, to run this solution and generate the sales forecasts, you need to register an account in Google Cloud. Then you have to create a new project, enable the BigQuery service to your account and configure your credentials.

Forecast Results

A report with the latest forecast results can be found at: Iowa Liquor Sales Forecast Report

Code Walkthrough

You can find a step-by-step walkthrough of the entire solution, including the data extraction, feature engineering, and transformation, model training and evaluation, as well as forecasting future sales at: notebooks/Walkthrough.ipynb

Pipelines

The pipelines folder contains scripts that can be used as entrypoints to perform several tasks related to the solution.

Additional Information

Docker Container

The Dockerfile defines the Docker container configuration to replicate the environment used to develop and run the forecasting model. By using this Docker container, you can ensure that the code runs consistently across different environments.

To build and run the Docker container, you can use the following commands:

  • Build the Docker image:

    docker build -t iowa-liquor-sales-forecast .
    
  • Run the Docker container:

    docker run -it --rm iowa-liquor-sales-forecast
    

Environment Variables

The solution relies on a few environment variables that need to be set up for proper operation. These include:

  • GOOGLE_APPLICATION_CREDENTIALS: Path to the JSON file that contains your Google Cloud service account credentials.

  • PROJECT_ID: The ID of your Google Cloud project.

  • DATASET_ID: The ID of the BigQuery dataset where the data is stored.

You can set these environment variables in your shell or define them in a .env file, which will be automatically loaded when running the Docker container or scripts.

Testing

The repository includes a suite of tests to ensure that the code behaves as expected. You can run the tests using pytest:

# Run tests
pytest tests/

Continuous Integration (CI)

This repository is set up with a Continuous Integration (CI) pipeline using GitHub Actions. The CI pipeline is configured to run the tests automatically whenever code is pushed to the repository or a pull request is created. This helps to ensure that new changes do not break existing functionality. It also contains a pipeline that recreates the documentation for the project and generates a new release of the documentation on GitHub Pages.

Here’s the list of currently available pipelines for the project:

  • deploy-docs.yml: deploy documentation to GitHub Pages.

  • test-code.yml: run the unit-tests from the tests directory and generate a test coverage report for the project.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Codebase Static Test Results

The iowa_forecast package received the following pylint scores:

  • iowa_forecast/__init__.py: 10.0

  • iowa_forecast/models_configs.py: 10.0

  • iowa_forecast/ml_train.py: 10.0

  • iowa_forecast/plots.py: 9.8

  • iowa_forecast/utils.py: 8.99

  • iowa_forecast/load_data.py: 9.31

  • iowa_forecast/ml_eval.py: 8.41

  • Average Score: 9.50

Modules

Contents:

Indices and tables