Publication

Abstract : Healthcare companies are in an endless state of flux. They are underneath massive stress to predict health concerns of clients and to create excess premium holders which will at the same time diminish the cost. Patient's readmission is often costly and shows shortfalls in the healthcare organizations. The cost of readmitted patients goes beyond 250 million dollars every year nationwide. Several healthcare agencies have started adopting Data Mining and Predictive Analysis. Predictive analysis involves various statistical techniques from modeling, machine learning, and data mining that breaks down past and present realism to forecasts the future. Henceforth, this paper is intended to propose a technique combining Apache Spark and deep learning based stacked ensemble method as a hybrid approach for predicting the readmission possibilities. Paper also focuses upon risk vindication strategies to predict patients with readmission possibility. This is implemented by considering medical data and estimating risk related using stacked machine learning techniques. With the application of such a framework that can satisfactorily classify the patient with readmission chance will help the healthcare companies to bestow top quality on healthcare systems. This technique helps achieve higher predictive accuracy of 90.69 % and RMSE score of 0.2521. Our empirical investigation demonstrates that this approach is helpful and can profit future research in the healthcare industry.