skope-rules

1.0.1 verified Mon Apr 27 auth: no python

skope-rules is a Python library for interpretable machine learning. It extracts decision rules from tree-based models to provide transparent predictions. Last version: 1.0.1 (released 2023). Low release cadence.

pip install skope-rules

Common errors

error ImportError: No module named 'skope_rules' ↓

cause Incorrect import path; module is installed as 'skrules'.

fix

Run 'from skrules import SkopeRules' instead.

error AttributeError: 'SkopeRules' object has no attribute 'predict_proba' ↓

cause skope-rules does not provide predict_proba by default.

fix

Use clf.predict() for class labels, or check the documentation for alternatives.

Warnings

gotcha The import path uses 'skrules' not 'skope_rules'. Using 'from skope_rules import SkopeRules' will raise ImportError. ↓

fix Use 'from skrules import SkopeRules'.

gotcha SkopeRules.score() returns a list of scores (one per class) for multi-class problems, not a single accuracy. This differs from sklearn convention. ↓

fix Use clf.score(X_test, y_test) which returns per-class scores. For overall accuracy, compute manually: (clf.predict(X_test) == y_test).mean()

deprecated Parameter 'max_depth_duplication' is deprecated in favor of 'max_depth' in future versions? Not officially deprecated yet, but be cautious with parameter names. ↓

fix Use 'max_depth' if available, else 'max_depth_duplication'.

Imports

SkopeRules

wrong

from skope_rules import SkopeRules

correct

from skrules import SkopeRules

Correct import uses 'skrules' not 'skope_rules'

Quickstart

Basic usage: train SkopeRules classifier on Iris dataset.

from skrules import SkopeRules
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split

X, y = load_iris(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

clf = SkopeRules(max_depth_duplication=2, n_estimators=10, random_state=42)
clf.fit(X_train, y_train)
scores = clf.score(X_test, y_test)
print(scores)