Modeling of diabetes mellitus risk based on consumption of salt, sugar, and fat factors using local linear estimator

Abstract

Diabetes Mellitus is a major health problem in the world. Diabetes Mellitus, commonly known as “the silent killer”, affects many of the body's systems and even leads to other serious diseases. The data from global studies showed that the number of people with Diabetes Mellitus in 2011 reached 366 million from all over the world. As a non- communicable disease, the prevalence of diabetes rises every year. Unhealthy eating habits, such as the consumption of salt, sugar and an excessive amount of fats, is one of the inflicting factors of this disease. For predicting diabetes mellitus risk based on salt, sugar and fat consumptions, we need to build a model. In statistical analysis, there are two approaches for estimating the model, i.e., parametric and nonparametric regression model. A local linear estimator is one of the estimators in nonparametric regression model that the advantages of this estimator can estimate the function at each point such that the model closes to the real pattern, and also no need large data to estimate the model. In this paper, we estimate the diabetes mellitus risk model based on salt, sugar and fat consumptions by using local linear estimator and compare it with logistic parametric regression approach. The result of this study, we get classification accuracies of diabetes mellitus risk based on salt, sugar and fat consumptions of 94.28% by using local linear estimator and of 80% by using parametric logistic regression. It means that nonparametric regression model approach by using local linear estimator is better than parametric logistic regression model approach.