Gradient Boosting Based Prediction Method for Patient Death in Hospital Treatment

摘要

Patient death in hospital is a concerned topic that is vital for both patients and hospital. The prediction of patient death in hospital is a classification process with extreme label imbalance problem which seriously affects the prediction effect of general classification model. In this paper, we use an ensemble learning method to predict patient death, the coordination of base classifiers in ensemble model can alleviate this imbalance. Patient measurement data, disease data and treatment data are used as inputs of the model, and whether the patient is died in hospital is estimated. From several comparison experiments, we evaluated several machine learning methods for patient mortality, Gradient Boosting based ensemble method outperforms other methods in AUC and other evaluation criteria, the highest AUC achieved by Gradient Boosting Classifier is 0.846. Finally, we proposed several future work based on our research.