A comparison of Machine Learning Methods to predict Hospital Readmission of Diabetic Patient

Abstract

Diabetes is a chronic disease whereby blood glucose is not metabolized in the body. Electronic health records (EHRs) for each individual or a population have become important to standing developing trends of diseases. Machine/Deep Learning helps provide accurate predictions higher than actual assessments. The main problem that we are trying to apply Machine/Deep learning model and using EHRs that combines the strength of a machine learning model with various features and Hyper-parameter optimization or tuning. The Hyper-parameter optimization uses the random search optimization which minimizes a predefined loss function on given independent data. The evaluation on the method comparisons indicated that Machine/Deep Learning models (Logistic Regression, Artificial Neural Network, Naïve Bayesian Classifier, Support Vector Machine and XGBoost) has improved results compared to the majority of previous models increasing the ratio of metrics (Accuracy, Recall, F1 and AUC score) on the same public dataset that is reprocessed. This shows that the proposed XGBoost model implemented in Amazon SageMaker (Amazon SageMaker was a Cloud Computing service) has the best performance evaluation results. This work is also one of the contributions to the global economic recovery in general and the reduction of medical equipment supply for the care and treatment of diabetics in particular during the Covid-19 pandemic.