A Predictive Analytic Multi-Model Approach for Estimating the Risk of Diabetic Forecast using Machine Learning Techniques

Abstract

The realm of disease monitoring and analysis plays an important role in day-to-day human life. Analysing and estimating the risk for a specified disease becomes challenging from region to region. This research concentrates towards the development of a multi-model approach for estimating the diabetic risk factors using machine learning techniques. Different sorts of practices such as Decision trees, Support Classifiers, Random tree Forest and Neural Network (FF) has been used for model development and risk estimation. The result analysis shows that hyperparameter tuning with signified estimators and finite states in random forest provides highest accuracy level. Statistical analysis has also been made using spearman rank correlation analysis. The results proved that risk factors pertaining to PPG, FPG and MBG with a correlation value of about 0.79 is found to be significant and correlated. Higher level of R-square value is observed among the risk factors PPG and FPG respectively. The interpretation and evaluation have made in accordance with the medical experts along with the preparation of dirt chart advisor for significant varied analysis.