Thirty-day Re-Hospitalization Rate Prediction of Diabetic Patients

Abstract

Diabetes is a serious global health problem, and re-hospitalization is usually associated with increased mortality and excessive medical burden. With the increasing cost of diabetes to the health care system, re-hospitalization is recommended as a goal to reduce health care costs. This paper aims to use data mining technology to accurately predict the 30-day re-hospitalization of diabetic patients. We use the data set from UCI machine learning repository, preprocessing, use feature reduction method to find out the classification results of re-hospitalization, and then use frequent set and Apriori algorithm to find the association rules between diabetes mellitus patients and re-hospitalization related variables. The experimental results show that the recursive feature reduction method is effective in combined with SVM can get a better prediction accuracy.