Hospital Length of Stay Prediction with Ensemble Methods in Machine Learning

The primary mission of hospitals is to meet the demand for care by moving patients through care pathways while improving coordination of care, patient safety, and health outcomes. The patient length of stay (LOS) in a hospital is used to gauge the efficiency of a healthcare facility. The goal of this study is to develop a machine learning system to predict LOS of patients at the admission phase, with initial diagnosis and test results. Classification (including deep learning) models are implemented and tested. The NY Hospital Inpatient Discharge in 2015 Dataset is used for model training and validation. Results are compared and discussed. The imbalanced nature of the dataset is discussed and addressed as well.