deboost

deboost implements weighted distance ensembling to aggregate predictions from multiple machine learning models for regression and classification, leveraging scikit-learn models and preprocessing functions.

Key Features:

• Weighted Distance Ensembling: Aggregates model outputs by computing and using weighted distances between predictions to inform ensemble weighting.
• Model Flexibility: Accepts any model that implements a predict method, including standard scikit-learn estimators.
• scikit-learn Integration: Leverages scikit-learn default models and data preprocessing functions as foundational components.
• Task Support: Applicable to both regression and classification tasks.
• Python Implementation: Provided as a Python library for machine learning ensembling.

Scientific Applications:

• Regression Ensembling: Refines ensemble predictions for regression problems by weighting inter-model prediction distances.
• Classification Ensembling: Produces consensus predictions for classification by aggregating model outputs based on weighted distances.
• High-Precision Predictive Studies: Enhances predictive accuracy and robustness in scientific analyses that benefit from ensemble methods.

Methodology:

Aggregates predictions from multiple models by computing weighted distances between model predictions and using those distances to weight the ensemble; leverages scikit-learn estimators and preprocessing and accepts models with a predict method.