Exploring the Relationship Between Smoothing Techniques and Model Interpretability in Nonlinear Regression Models

Posted: May 20, 2020

• Austin Rivera
• Brandon West
• Caleb Ross

Abstract

The proliferation of complex nonlinear regression models in modern data science has created an inherent tension between model performance and interpretability. While smoothing techniques have long been established as essential tools for regularization and noise reduction in statistical modeling, their impact on model interpretability remains a critically understudied aspect of the machine learning paradigm. Traditional approaches to smoothing primarily focus on optimizing predictive accuracy and generalization performance, often neglecting the consequences for model transparency and explanatory power. This research addresses this significant gap by systematically investigating how different smoothing methodologies influence various dimensions of interpretability in nonlinear regression frameworks.