Welcome to stable-cart’s documentation!¶

stable-cart provides individual decision trees with stability-focused modifications to the tree-building process. All trees follow the familiar scikit-learn API while incorporating advanced stability features.

Key Features¶

🌳 Individual Decision Trees: Single trees (not ensembles) with stability-focused modifications

🎯 Stability Mechanisms: Multiple approaches including data partitioning, consensus, and bootstrap methods

📊 sklearn Compatible: Works seamlessly with pipelines, cross-validation, and grid search

🔬 Analysis Tools: Bootstrap variance measurement for evaluating prediction stability

Quick Start¶

from stable_cart import LessGreedyHybridTree
from sklearn.datasets import make_classification

# Generate sample data
X, y = make_classification(n_samples=1000, n_features=10, random_state=42)

# Create and train a stable tree
tree = LessGreedyHybridTree(
    task='classification',
    max_depth=6,
    min_samples_leaf=2,
    split_frac=0.9,
    val_frac=0.05,
    est_frac=0.05,
    random_state=42
)
tree.fit(X, y)
predictions = tree.predict(X)

See the tree_estimators section for a complete list of available estimators and the evaluation_functions for assessing model stability.

Documentation¶

API Reference:

Examples:

Links: