Machine learning applications towards risk prediction and cost forecasting in healthcare
The United States spends a considerable amount on healthcare and health related expenditures. A sizeable portion of these costs are created by the performing of and recovery from surgery. In this report machine learning methodologies are used to predict the potential risk an individual undergoing elective spinal surgery has for a high-cost recovery. Two models are built in this report, a time-series model to forecast future costs forward over an unknown time horizon, and a multi-class classifier to create the risk predictions. The results of these different model architectures are compared to one another, and the best are then stacked to create a final ensemble learner.