Using tidymodels to create a viable ML model for binary classification and explaining variable contribution for individual prediction scores
This project shows a general approach to generating a XGBoost binary classification machine learning (ML) model using the 'tidymodels' package. A simulated dataset is generated to reflect anthropometric/biomarker z-scores for healthy and sick individuals, representing the current classification dichotomy. The aim is to produce a viable classification model, basic model documentation and explain variable contribution in individual prediction scores by 'break-down' analyses.