A COMPREHENSIVE ANALYSIS OF MACHINE LEARNING TECHNIQUES FOR INCESSANT PREDICTION OF DIABETES MELLITUS

People in every country including Indians is severely affected by diabetes. Based on its importance diabetes researchers across the world are working towards thousands of different research goals. In this paper prediction problem of the diabetes patients' early readmission is considered. The main aim of the work is to help the doctors and patients to predict the hospital readmission using a better performing machine learning algorithm. Given the diabetes dataset as input, different machine learning mechanisms are applied on Pima Indian Diabetes Dataset for comparison and obtaining the best performing algorithm. The Machine learning algorithms considered in this work are Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Adaptive Boosting (AB) and Gradient Boosting (GB). All these algorithms are applied on diabetes dataset from 130 hospitals from 1999-2008 provided by UCI machine learning repository. Fourty six attributes of this diabetes dataset are taken as input variables and hospital readmitted attribute is taken as output variable. All these algorithms are comprehensively analyzed based on the precision, recall and f1-score. This analysis could be used to select the suitable prediction algorithm. Then the task is to predict whether a diabetes patient is early readmitted with diabetes risks or not using the optimal algorithm. Finally the prediction could be used by doctors, patients and hospital authorities to monitor the diabetic patients.