Adaptive Optimization for Prediction with Missing Data

Dimitris Bertsimas, Arthur Delarue, Jean Pauphilet

When training predictive models on data with missing entries, the most widely used and versatile approach is a pipeline technique where we first impute missing entries and then com…