Applying Deep Belief Network for Predicting the Risk of Diabetes Complications

The acquisition of knowledge and significant outcomes from complex, multi-dimensional, and diversified medical information continues to be a significant obstacle to providing better healthcare. Every day, new types of clinical information emerge as a result of the overwhelming number of people who must undergo medical tests. Among these are text, images, and data from sensors and electronic health records (EHRs). There is a lack of organization, diversity, good commentary, and predictability in these. In order to sift through and draw conclusions regarding these medical datasets, effective data analytics methods are required. A Deep Belief Network (DBN) model for Type 2 Diabetes Mellitus (T2D) issue prediction is suggested in the subsequent phase of the study. The model sheds light on the prevalence of diabetes and its potential consequences. We compare the accuracy, precision, and recall of DBN-based analytical data models to those of Support Vector Machines (SVMs), Random Forests (RFs), K Nearest Neighbours (KNNs), Artificial Neural Network (ANNs), and Convolutional Neural Networks (CNNs). Throughout training, the suggested approach accurately establishes the widespread nature of risk with a 97.39% accuracy rate, but during examination.