MLwrap: Simplifying Machine Learning Workflows in R
TL;DR
A unified, minimalistic interface covering all predictive modeling stages: data preprocessing, model construction, hyperparameter optimization, evaluation, and sensitivity analysis, that empowers analysts and researchers to efficiently extract actionable insights with transparency.
Abstract
MLwrap is an R package that streamlines machine learning (ML) workflows, making them accessible, efficient, and reproducible, especially within the Knowledge Discovery in Databases process. It offers a unified, minimalistic interface covering all predictive modeling stages: data preprocessing, model construction, hyperparameter optimization, evaluation, and sensitivity analysis. MLwrap supports popular algorithms like Neural Networks, Support Vector Machines, Random Forests, and XGBoost Decision Trees. The workflow is organized into four core functions: preprocessing(), build_model(), fine_tuning(), and sensitivity_analysis(). These functions encapsulate the entire ML pipeline into a single, reproducible analysis object. Demonstrated via regression and classification examples, MLwrap simplifies complex tasks, providing interpretable results. It empowers analysts and researchers, particularly in health and social sciences, to efficiently extract actionable insights with transparency.