MAIF / shapash

🔅 Shapash: User-friendly Explainability and Interpretability to Develop Reliable and Transparent Machine Learning Models
https://maif.github.io/shapash/
Apache License 2.0
2.73k stars 334 forks source link

Lack of Customization for Graph Size and Incorrect Color Display in Shapash Explainability Quality Graphs #610

Closed guillaume-vignal closed 4 days ago

guillaume-vignal commented 4 days ago

Description:
The current implementation of the Shapash library’s explainability quality graphs has limitations and a few issues affecting usability and visual clarity:

  1. Fixed Graph Size:

    • Problem: There is no way to adjust the height and width of explainability quality graphs, which restricts flexibility when visualizing data with a large number of features or when specific layout adjustments are required.
    • Expected Improvement: Introducing customizable height and width parameters to allow users to set the dimensions of the graphs as needed.
  2. Title Positioning:

    • Problem: The graph title appears misaligned, making it harder to read and affecting the overall presentation.
    • Expected Improvement: Adjusting the title alignment, possibly with a line break or formatting changes, to ensure it is properly positioned for better readability.
  3. Incorrect Color Palette Selection:

    • Problem: There is an issue with the selection of color palettes in the explainability quality graphs, leading to incorrect or inconsistent color displays. This can make it difficult to interpret the graphs accurately.
    • Expected Improvement: Correcting the palette selection process to ensure that the chosen colors are applied consistently, improving the graph’s clarity and usability.

Suggested Solution:
To address these issues, it would be beneficial to:

Context:
These updates would provide users with greater control over the appearance of explainability quality graphs and resolve the current visual inconsistencies, enhancing the usability and visual quality of Shapash's explainability tools.