Add plot_decision_regions_3d for stacked 3D decision regions #488

[mariam851]

Hi @rasbt,

I have implemented the plot_decision_regions_3d function to address the request in issue #488. Following your suggestion in the discussion, I have created this as a standalone function to maintain the modularity of the library and avoid over-complicating the existing 2D plotting logic.

Key Features:

• Stacked 3D Visualization: The function allows users to visualize how decision boundaries shift across different values of a third feature by "stacking" 2D slices at specified $Z$ offsets.
• Smart Grid Generation: It dynamically generates a mesh grid based on the primary features and performs predictions across the specified $Z$ slices.
• Integrated Scatter Plotting: Includes an option to plot the original 3D data points (scatter_points=True) to provide context for the decision surfaces.
• Polymorphic Design: Works with any scikit-learn compatible classifier that implements a predict method.

Example Output:

I have tested the implementation with a linear SVM on a 3-feature dataset. As seen in the image below, the function clearly illustrates the decision surfaces at various levels of the third feature:

I believe this addition will be very helpful for users trying to interpret higher-dimensional models. Looking forward to your feedback!

Fixes #488

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB