classtrees 1.1.0

Python package implementing classification trees.

classtrees

classtrees is a lightweight Python package that implements fast classification trees and random forest. It combines a NumPy-friendly Python API with a compiled C backend to deliver efficient training and prediction for binary and multiclass classification.

Key Features

• Decision tree classifier with configurable tree height, split criteria, and feature subsampling
• Random forest ensemble with bootstrap aggregation and support for parallel training
• Predict class labels and class probabilities
• Built as a C extension for performance
• Designed for NumPy arrays and scientific workflows

Installation

Requirements

• Python 3.10 or newer
• NumPy
• C compiler with OpenMP support (required to build the C extension)

Install from PyPI

python -m pip install classtrees

Install from source

python -m pip install .

Quick Start

import numpy as np
from classtrees import ClassTree, RandomForest

# training data
X = np.random.randn(100, 5)
y = (X[:, 0] > 0).astype(np.int64)

# single decision tree
tree = ClassTree(max_height=5, random_state=42)
tree.fit(X, y)
y_pred = tree.predict(X)
proba = tree.predict_proba(X)

# random forest
forest = RandomForest(n_estimators=50, max_height=8, random_state=42, n_jobs=1)
forest.fit(X, y)
forest_pred = forest.predict(X)
forest_proba = forest.predict_proba(X)

API

ClassTree

A single decision tree classifier.

Constructor arguments:

• impurity: 'gini' or 'entropy' (default: 'gini')
• max_height: maximum tree height, or None to grow until no valid split remains
• min_samples_split: minimum samples required to split a node (default: 2)
• min_samples_leaf: minimum samples required in each leaf node (default: 1)
• max_features: number of features considered for splits, or None to use all features
• random_state: random seed for reproducible training

Methods:

• fit(X, y): train the classifier on feature matrix X and labels y
• predict(X): return predicted class labels for input samples
• predict_proba(X): return predicted class probabilities

RandomForest

An ensemble of decision trees with bootstrap sampling.

Constructor arguments:

• n_estimators: number of trees in the forest (default: 100)
• impurity, max_height, min_samples_split, min_samples_leaf, max_features, random_state: same as ClassTree
• n_jobs: number of worker threads for training and prediction (default: 1). Use -1 to enable all available cores.

Methods:

• fit(X, y): train the forest ensemble
• predict(X): return ensemble class predictions
• predict_proba(X): return averaged class probabilities from all trees

Documentation

Full documentation is available in the project docs/ folder and online at:

• https://kubek22.github.io/classtrees/

License

This project is licensed under the BSD-3-Clause license. See LICENSE for details.

Project Links

• Homepage: https://github.com/kubek22/classtrees
• Repository: https://github.com/kubek22/classtrees