causal-drift 0.1.0

A causal feature selection (Causal DRIFT: Causal Dimensionality Reduction via Inference of Feature Treatments) library using residual-based ATE estimation.

deepCausal

deepCausal is a Python library for causal feature ranking and selection based on residual treatment effect estimation.
It uses double machine learning and residualization techniques to estimate the causal impact (ATE) of each feature on a continuous or categorical outcome.

🔍 Key Features

• Supports both continuous and categorical outcomes
• Uses residual-on-residual regression for causal interpretation
• Identifies confounders automatically
• Ranks features by causal strength, not just correlation

🚀 Usage

from deepCausal import DeepCausal import pandas as pd import numpy as np

df = pd.DataFrame({ 'X1': np.random.randn(100), 'X2': np.random.rand(100), 'X3': np.random.randn(100), 'Y': np.random.randn(100) })

X = df[['X1', 'X2', 'X3']] y = df['Y']

model = DeepCausal() model.fit(X, y, outcome_type='continuous')

print(model.get_feature_ate())

📄 License This project is licensed under the MIT License.

🧪 Disclaimer This method is experimental and currently under research evaluation. Feedback and pull requests are welcome!