🛠 scikit-mtr: Multi-Task Regression in Python

scikit-mtr provides a framework for multi-task regression using popular regression algorithms and introduces a stacking method to combine different regressors for enhanced performance.

🌟 Features

• 📊 Support for multiple regressors: Decision Trees, Linear Regression, Random Forests, Extra Trees, MLP, and MOKP.
• 🏆 Stacking method to combine predictions from various regressors.

📦 Installation

pip install scikit-mtr

⚡ Quick Start

Get up and running with scikit-mtr in a flash:

from sklearn.model_selection import train_test_split

from scikit_mtr.multi_output_tools import multi_output_regressor, load_data_by_sklearn

params = {
    'regressor': 'LR',
    'dataset': 41467,
    'random_seed': 0
}

# Load and split data
X, y = load_data_by_sklearn(params['dataset'])
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=params['random_seed'])

# Initialize regressor
model = multi_output_regressor(params['regressor'])
model.fit(X_train, y_train)

# 🎯 Make predictions
y_pred_test = model.predict(X_test)

📚 Using Stacking

scikit-mtr introduces a powerful stacking method where base regressors' predictions serve as features for meta-regressors:

from scikit_mtr.multi_output_stacking import MultiTargetRegressorStacking

base_regressors = [LinearRegression() for _ in range(y_train.shape[1])]
meta_regressors = [LinearRegression() for _ in range(y_train.shape[1])]

stacker = MultiTargetRegressorStacking(base_regressors, meta_regressors)
stacker.fit(X_train, y_train)
stacked_predictions = stacker.predict(X_test)
print(stacked_predictions)

🧪 Supported Regressors

Currently, the following regressors are in our arsenal:

• DT: 🌲 Decision Tree
• LR: 📈 Linear Regression
• RF: 🌳 Random Forest
• ET: 🍃 Extra Trees
• MLP: 🧠 Multi-layer Perceptron
• MOKP: 🔮 MOKP Regressor