Prediction of Diabetes Readmission using Machine Learning
Predictive analytics and machine learning have quickly become some of the most talked about topics in healthcare analytics. Machine Learning and its applications have a proven track record in many industries. From these previous success cases, valuable lessons can be learned to utilise predictive techniques in healthcare analytics for patient care improvement, hospital administration, chronic disease management, and supply chain efficiencies. Patient readmission in hospitals—of which the ones associated with diabetes are most regular—has become like an epidemic and its data has become a primary source of information for finding areas of improvement in healthcare. In this project we use binary classification algorithms on diabetic patient data from the US, extracted from the UCI Machine Learning Repository, to predict patients’ chances of readmission within 30 days and find which factors could substantially increase that risk. Further, we attempt to reduce the number of features by a) doctor consultation and b) filter methods of feature selection applied on the algorithms obtained in the primary module, and thereby improve the resulting accuracy.