A STUDY OF JUSTIFICATION ON JUPYTER NOTEBOOK QUALITY & FAIRNESS

Abstract

Computational notebooks using the Jupyter Platform have become increasingly popular among data scientists and machine learning engineers. However, due to the diversity of tools and languages, Notebook developers face reproducibility challenges. Limited research has been conducted on verifying the quality of the notebook, and only a few studies have established best practices for notebook development. Thus, further study is needed to transform those conceptual best practices into concrete, actionable quality checks to ensure the quality & fairness of the notebook. This report aims to improve the quality of notebooks by investigating the possibility of using the Justification Diagram Language to convert best practices into tangible steps of quality checks that users can execute within the Continuous Integration and Deployment (CI/CD) pipeline. The report will focus on justifying 12 notebooks’ best practices collected from existing studies using Justification Diagrams, and it will then map these diagrams into practical steps that can be run in GitHub Actions, demonstrating their effectiveness in real-life scenarios and thus enriching the quality of the notebooks.