Author/Authors :
farajollahi, boshra iran university of medical sciences - school of health management and information sciences - department of health information management, n management, school of health management and information sciences, iran university of medical sciences, tehran, iran , mehmannavaz, maysam doornama company - data science lab, tehran, iran , mehrjoo, hafez doornama company - data science lab, tehran, iran , moghbeli, fateme varastegan institute of medical sciences, mashhad, iran , sayadi, mohammad javad iran university of medical sciences - school of health management and information sciences - department of health information management, tehran, iran
Abstract :
introduction: diabetes is a chronic disease associated with abnormal high levels of glucose in the blood. diabetes make many kinds of complications, which also leads to a high rate of repeated admission of patients with diabetes. the goal of this study is to predict hospital readmission of diabetic patients with machine learning techniques. material and methods: the data used in the study are data obtained from the uci machine learning repository about diabetic patients. the dataset used contains 100,000 instances and it include 55 features from 130 hospitals in the united states for 10 years. results: this article gets results from the final stages of evaluation. in this evaluation process, compared the performance of decision tree, random forest, xgboost, kneighbors, adaboost and deep neural network with accuracy. conclusion: the number of selected features by pcabased feature selection method improve the predictive performance based on accuracy of deep learning and most machine learning models for predicting readmission. the improvement of machine learning models depended on the specific choice of the prediction model, number of selected features, and “k” for kfold validation.
