Improving Accuracy for Diabetes Mellitus Prediction Using Data Pre-Processing and Various New Learning Models

Abstract

Data mining in medical data has successfully converted raw material into useful information. This information helps the medical experts in improving the diagnosis and treatment of diseases. Type II Diabetes Mellitus is one of silent killer diseases worldwide. According to World Health Organization, 346 million people are suffering from diabetes worldwide. Diagnosis or prediction of Diabetes is done through various data mining technique such as association, classification, clustering and pattern recognition. The study led to related open issues of identifying the need of a relation between the major factors that lead to the development of diabetes. This is possible by mining patterns found between the independent and dependant variable in the dataset. This paper compares classification accuracies of various machine learning models. Objective of paper is to find whether a person has diabetes or not and what features are highly responsible for diabetes. As due to its continuously increasing occurrences more and more families are influenced by diabetes mellitus. Most diabetic people know little about their health. In this study, we have proposed novel model on data mining techniques for predicting type 2 diabetes mellitus. Diabetes often referred to by doctors as metabolic disease in which the person has high blood glucose (blood sugar), because of inadequate insulin production.