Add SHAP-like beeswarm plot visualization

[yasumorishima]

Summary

Add plot_beeswarm visualization package that creates SHAP-style beeswarm plots for Optuna studies.

Closes #116

Motivation

Requested in optuna/optuna#4987 and redirected to OptunaHub in #116. Beeswarm plots reveal monotonic relationships between hyperparameter values and objective function values, complementing existing visualizations like slice plot and parallel coordinate plot.

What it does

• X-axis: Objective function value
• Y-axis: Parameters sorted by importance (most important at top)
• Color: Normalized parameter value (blue=low, red=high)
• Vertical spread: Density-based jitter — wider where trials are concentrated

Implementation details

• Histogram-based density estimation for beeswarm jitter (no scipy dependency)
• Parameter sorting: fANOVA (via optuna.importance) with Spearman rank correlation fallback when sklearn is unavailable
• Dependencies: only matplotlib and numpy (already common in Optuna ecosystem)

Files

package/visualization/plot_beeswarm/
├── __init__.py    # plot_beeswarm() function
├── README.md      # Documentation with frontmatter
├── example.py     # Minimal working example
├── LICENSE        # MIT
└── images/
    └── beeswarm.png

Checklist

• Pre-commit hooks pass (ruff, ruff-format, mypy, mdformat)
• Minimal working example provided
• Sample image included
• I agree to the contributor agreements

[c-bata]

LGTM! Thank you for your contribution.

[y0z]

Published: https://hub.optuna.org/visualization/plot_beeswarm/