featureexpand 0.4.0

A powerful Python library for automatic feature engineering using logical formula simplification

FeatureExpand

FeatureExpand is a powerful Python library designed to enhance your datasets through automatic feature engineering using logical formula simplification. Whether you're working on machine learning, data analysis, or any data-driven application, FeatureExpand helps you extract maximum value from your data by generating optimized boolean and continuous features.

✨ Features

• Automatic Feature Engineering: Generates new features using logical formulas optimized by the Exactor API
• Scikit-learn Compatible: Fully compatible with scikit-learn pipelines and transformers
• Boolean & Continuous Features: Supports both boolean logic features and continuous relaxations
• Model Booster: LinearModelBooster automatically selects the best model among linear, polynomial, and logic-guided regression
• Smart Caching: Cached API calls for improved performance
• Flexible Configuration: Customizable precision, depth, and feature generation strategies

🚀 Quick Start

Installation

pip install featureexpand

Basic Usage

import numpy as np
import pandas as pd
from featureexpand import FeatureExpander, LinearModelBooster
from sklearn.model_selection import train_test_split
from sklearn.metrics import r2_score

# Create sample data
X = pd.DataFrame({
    'feature1': np.random.rand(100),
    'feature2': np.random.rand(100),
})
y = (X['feature1'] > 0.5) & (X['feature2'] < 0.5)  # Boolean logic target

# Split data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Use FeatureExpander
expander = FeatureExpander(
    token="your_exactor_api_token",  # Get from https://www.booloptimizer.com
    deep=2,  # Bits per feature
    environment="PROD"
)

# Fit and transform
X_train_expanded = expander.fit_transform(X_train, y_train)
X_test_expanded = expander.transform(X_test)

print(f"Original features: {X_train.shape[1]}")
print(f"Expanded features: {X_train_expanded.shape[1]}")
print(f"Generated formula: {expander.formula_string_}")

Using LinearModelBooster

from featureexpand import LinearModelBooster

# Automatically selects best model (Linear, Polynomial, or Logic-Guided)
booster = LinearModelBooster(
    token="your_exactor_api_token",
    poly_degrees=[2, 3],
    logic_depths=[1, 2],
    cv=5,
    scoring='r2'
)

# Fit and predict
booster.fit(X_train, y_train)
y_pred = booster.predict(X_test)

# View results
booster.summary()
print(f"Best model: {booster.best_model_name_}")
print(f"Best score: {booster.best_score_:.4f}")

📚 Documentation

For comprehensive documentation, including tutorials, API reference, and examples, visit:

• Documentation: https://featureexpand.readthedocs.io
• Examples: See the examples/ directory in the repository

🔑 API Token

FeatureExpand uses the Exactor API for boolean formula optimization. You can:

1. Get a free token at https://www.booloptimizer.com
2. Set it as an environment variable: export EXACTOR_API_TOKEN=your_token
3. Pass it directly: FeatureExpander(token="your_token")

💡 Use Cases

• Machine Learning: Enhance model performance with engineered features
• Binary Classification: Extract logical patterns from data
• Feature Discovery: Automatically discover feature interactions
• Model Interpretation: Understand data relationships through boolean formulas
• Trading Signals: Generate forex/stock trading signals (see examples/forex/)

🛠️ Requirements

• Python >= 3.8
• NumPy >= 1.19.0
• Pandas >= 1.1.0
• Scikit-learn >= 0.24.0
• Requests >= 2.25.0

📦 What's Included

• FeatureExpander: Core transformer for automatic feature generation
• LinearModelBooster: Automatic model selection with feature expansion
• Neural utilities: Neural network guidance for feature engineering
• Utility functions: Encoding, formula conversion, and migration tools

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Powered by the Exactor API for boolean formula optimization
• Built with Scikit-learn
• Inspired by research in boolean function minimization and automatic feature engineering

📞 Contact

• Author: Juan Carlos Lopez Gonzalez
• Email: [e-mail]
• GitHub: https://gitlab.com/exactor/featureexpand.git
• Issues: https://gitlab.com/exactor/featureexpand.git/issues

---

Made with ❤️ for the ML community