Skip to content

khuyentran1401/reproducible-data-science

Repository files navigation

View on GitHub View Book

Build a Reproducible and Maintainable Data Science Project

As data science projects increase in requirements, such as reliability, maintainability, and scalability, the complexity of projects increases significantly.

Image by xkcd

Thus, having reproducible data science workflows ensures consistency in our results, making it easier to debug and maintain these projects.

This book introduces Python tools for developing efficient workflows for reproducible and maintainable data science projects. We introduce best practices and tools which enable data scientists to be able to adapt to the ever growing demand in complexity, while ensuring that their systems are reliable.

At the end of this book you will learn how to structure your project, effectively use parameters, loggers, and pipelines to be able to test, debug, and build reproducible results from your workflows.

What is Reproducibility?

If a data science project is reproducible, results obtained from the project should be achieved again with a high degree of reliability when the project is replicated by another person in another machine.

Image by xkcd.

What is Maintainability?

If a data science project is maintainable, others can debug, maintenance, and add more features to the project with ease without breaking the code.

Image by xkcd.

About The Author

Khuyen Tran wrote over 150 data science articles with 100k+ views per month on Towards Data Science. She also wrote 500+ daily data science tips at Data Science Simplified. Her current mission is to make open-source more accessible to the data science community.