BASED ON HYBRID CNN-SVM DIABETES PATIENTS' PREDICTIVE HOSPITAL READMISSION MODEL

Abstract

A multitude of factors, such as an ageing population, an increasing reliance on technology, and increased patient expectations, present challenges for today's healthcare systems. These modifications have led to the development of more patient- and value-driven healthcare systems. For many organizations, it is quite challenging to increase health care while reducing expenditures. The broad deployment of electronic medical records and improvements in processing capacity have greatly increased access to electronic health data (EMRs). Typical programs lack the processing power to effectively handle big data sets since there are so many data to process. An intelligent technology utilized for prediction, the combined Convolutional Neural Network (CNN) and Support Vector Machine (SVM) has produced encouraging results. The hinge loss function for combining CNN and SVM is discussed in this article. Feature engineering and the insertion of missing variables are both taken care of in the preprocessing steps. The main result of this study, which employs a CNN-SVM model, is an increase in readmission prediction accuracy to 78%.