MSR35 Data Pre-Processing Approaches in Predictive Machine Learning Observational Studies

An integral part of conducting research using machine learning (ML) approaches involves robust pre-treatment of features (or variables) before the development of ML-based predictive models to improve predictive precision. However, data pre-processing steps are not well described in healthcare literature. In this study, the data pre-processing steps used to develop ML predictive models including LASSO regression, Elastic Net Logistic (ELN) regression, Classification and Regression Tree (CART), Random Forest (RF), and Gradient Boost ML (GBML) on a RCT dataset will be described.