Evaluation of Deep Belief Network to Predict Hospital Readmission of Diabetic Patients

In the present research trend of data mining and machine learning, one of the most important subject is disease prediction especially diabetes prediction and control. Diabetes prediction is a subject of utmost importance due to its risks and side effects. By predicting the risks of the diabetes countries could save the most valuable human resources. The problem under consideration of this work is an effective prediction of diabetes patient's early readmission into a hospital. This problem is proposed to be solved in this paper by using a deep learning algorithm namely Deep belief network. The Pima Indian Diabetes Dataset is considered in this paper. The proposed algorithm alongside existing algorithms are implemented in R. Then the proposed algorithm is compared with the considered algorithms using evaluation measures like Precision, Accuracy Specificity, NPV and F1-score for ascertaining the effectiveness of proposed deep learning algorithm. Then the Deep belief network could be used to predict the possibility of a diabetes patient to be early readmitted. This information could be finally used by hospital managements to optimally utilize the hospital resources and also to continuously observe the patients.