This respository contains code for the project "Predicting hospital readmission of diabetic patients using ensemble learning"
Diabetes is a disorder in which there are high blood sugar levels. Therefore, the assessment of diabetes patients during hospitalization is of great importance. The goal of this research is to predict the probability that a diabetic patient is readmitted within 30 days of hospitalization. Predicting readmission could help to reduce cost of care, medical dispute and improve patients health and safety. The dataset used for this research was submitted by the Center for Clinical and Translational Research, Virginia Commonwealth University, and is available on the University of California Irvine Machine Learning Repository. In this paper, different bagging based ensemble models were created using logistic regression, naive bayes, random forest, K-nearest neighbours and extreme gradient boosting. As a result, the ensemble with logistic regression, extreme gradient boosting, random forest, and naive bayes provides the most accurate results with an AUC of 61.45%. Therefore, it is recommended to further improve this ensemble learning algorithm in order to improve the accuracy and AUC.