Deep Belief Neural Network Model for Prediction of Diabetes Mellitus

Abstract

Diabetes Mellitus is metabolic chronic disease in which blood glucose levels are too high. In India nearly 8.7% of population suffers from diabetes in age range from 20 to 70. Unidentified and untreated diabetes leads to so many health difficulties such as damage of heart, kidneys, eyes, nerves and blood vessels. There are already several methods exists to support clinical decision making but still need improvements to solve the issues and challenges. In this research work, deep belief network model is designed for providing computational intelligence for prediction of patient affected by diabetes mellitus with maximum accuracy. Pima Indians Diabetes Dataset is used to analyze and experiment this prediction model. Firstly, the dataset is pre-processed by applying normalization technique. Secondly, the prediction model using deep belief neural network is designed. At the end, an experimental results proved that the comparison of overall performance of deep belief networks method is better than familiar classifiers namely naïve Bayes, Decision Tree., Logistic Regression (LR), Random Forest (RF) and Support Vector Machine (SVM).