radtree 0.1.0

Radial visualization helpers for scikit-learn decision trees

radtree

radtree visualizes scikit-learn decision tree predictions as a radial graph. It is designed for quick exploratory inspection of class boundaries and prediction consistency on a sampled evaluation set.

Status

This project is maintained for exploratory decision-tree visualization workflows. Core public APIs (plot_radial, quick_fitted_tree, and package imports) are kept stable for notebook and script usage.

Installation

Local editable install (recommended for development)

python -m venv .venv
source .venv/bin/activate
python -m pip install -U pip
python -m pip install -e .

PyPI install (when published)

pip install radtree

If a PyPI release is not available yet, use editable install or GitHub source install.

Quickstart (<30 seconds)

from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier

import radtree

random_state = 42
X, y = datasets.load_iris(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.4, random_state=random_state
)

clf = DecisionTreeClassifier(random_state=random_state)
clf.fit(X_train, y_train)

fig, ax = radtree.plot_radial(
    clf,
    X=X_test,
    Y=y_test,
    smooth_d=8,
    l_alpha=0.2,
    l_width=1,
    random_state=random_state,
)

API

plot_radial

plot_radial(
    clf,
    X=None,
    Y=None,
    data=None,
    feature_cols=None,
    label_col=None,
    num_samples=100,
    levels=None,
    edges_labels=None,
    draw_labels=None,
    style="radplot",
    bbox="dark",
    cmap="pairs",
    tree_node_size=50,
    leaf_node_size=50,
    node_size=50,
    l_width=1,
    l_alpha=1,
    fig_res=72,
    save_img=False,
    img_res=300,
    png_transparent=True,
    spring=False,
    smooth_edges=False,
    smooth_d=None,
    smooth_res=50,
    random_state=None,
)

Core parameters:

• clf: fitted DecisionTreeClassifier.
• X/Y or data: input records and labels.
• num_samples: number of rows to render. Use None to use all rows.
• random_state: deterministic sampling seed.
• smooth_d: enables spline smoothing when set.
• save_img: saves output PNG to ./plots/.

Return value:

• (fig, ax) matplotlib figure and axes.

Additional helpers:

• quick_fitted_tree
• plot_pca
• plot_tsne
• plot_umap (requires optional dependency umap-learn)

Performance and limits

plot_radial uses num_samples=100 by default as a safe rendering baseline. Runtime grows with sample size, and grows faster when smoothing is enabled.

Practical guidance:

• Start with the default num_samples=100.
• Increase to 200+ only when needed.
• Keep smooth_d=None for faster previews.
• Use random_state for repeatable comparisons.
• For full-dataset runs (num_samples=None), prefer smaller evaluation subsets if rendering gets slow.

Examples

Notebooks and example datasets are available in notebooks/.

Third-party notices

Third-party dependency notices are listed in THIRD_PARTY_NOTICES.md.

Contributing

Please read CONTRIBUTING.md for local setup, quality checks, and pull request expectations.

License

MIT License (LICENSE).