國立臺南大學應用數學系

講   題: A Two-Stage Machine Learning Approach to Predict Heart Transplantation Survival Probabilities over Time with a Monotonic Probability Constraint 
演講者: Prof. Ying-Ju (Tessa) Chen, Department of Mathematics, University of Dayton 
時   間: 108/06/12(三) 下午 3:00-5:00 
地   點: 府城校區 格致樓C305室 
摘   要: Accurate prediction of graft survival after a heart transplant is an important, yet challenging problem because: (a) it is the only treatment option for patients with end-stage heart failure; (b) the availability of hearts from deceased donors is scarce; (c) it requires an estimation of the matching suitability of patient-donor based on their medical information; and (d) its success is affected by the patient’s adherence to strict medical instructions after transplant. In this study, we propose a two-stage approach for estimating the graft survival probabilities at 1-10 years post surgery. First, we estimate the survival probability at different time points using machine learning methods. Then, we calibrate these probabilities using isotonic regression. This study provides two general implications: (a) accurate predictions can be obtained using the two-stage methodology; and (b) using isotonic regression to calibrate machine learning models in survival analysis can lead to more informative results.