Machine Learning Regression (MLR) diagnostics

This module provides various tools to create and evaluate MLR models for arbitrary input variables.

Examples

• Constraining uncertainty in projected gross primary production (GPP) with machine learning: Use Gradient Boosted Regression Tree (GBRT) algorithm to constrain projected Gross Primary Production (GPP) in RCP 8.5 scenario using observations of process-based predictors.

Diagnostic scripts

• Evaluate residuals
• MLR main diagnostic
• Multi-model means (MMM)
• Plotting functionalities
• Postprocessing functionalities
• Preprocessing functionalities
• Rescale data with emergent constraints

Auxiliary scripts

• Auxiliary functions for MLR scripts
• Custom extensions of sklearn functionalities
• MLRModel base class
• Base class for Gradient Boosted Regression models
• Base class for Linear models

Available MLR models

• Gradient Boosted Regression Trees (sklearn implementation)
• Gradient Boosted Regression Trees (xgboost implementation)
• Gaussian Process Regression (sklearn implementation)
• Huber Regression
• Kernel Ridge Regression
• LASSO Regression
• LASSO Regression with built-in CV
• LASSO Regression (using Least-angle Regression algorithm) with built-in CV
• Linear Regression
• Random Forest Regression
• Ridge Regression
• Ridge Regression with built-in CV
• Support Vector Regression