Performance Evaluation of Data Mining and Neural Network Based Models For Diabetes Prediction

Abstract

Diabetes, often known as diabetes mellitus, is a disease that disrupts the body’s normal response to blood sugar. The pancreas releases insulin, which aids in the uptake of glucose from meals into cells to be used as fuel. Hyper-glycemia,or high blood sugar, is a typical result of uncontrolled diabetes and is associated with several significant health complications, most notably those affecting the nerves and blood vessels. Statistics indicate that in 2014, adults 18 and above had diabetes, and in 2019, diabetes was responsible for 1.5 million fatalities worldwide. Machine learning and deep learning predictive models have seen tremendous development throughout industries, including health care, making early diagnosis of diabetes a breeze. Many potentially fatal diseases, such as cancer, diabetes, heart disease, thyroid disease, etc., may be predicted or diagnosed with the use of machine learning classifiers. The treatment of chronic diabetes, one of the world’smost prevalent illnesses, might benefit greatly from improved diagnostic efficiency.Here, we examine the relative merits among several ML and DL approaches to the problem of early diabetic illness prediction. The fundamental purpose of this research is to organize and conduct out diabetes prognosis with several ML learning approaches and then analyze the results of these methods to determine which one is the most accurate classifier. In this work, we take a multifaceted approach to diabetes and its prediction by investigating a wide range of disease-related characteristics. We use the classic Dataset Based on PIDD, and we apply several Machine Learning and Deep learning classifiers to it, including Random Forest (RF), Logisticregression (LR), Support Vector Machine, Artificial Neural Network (ANN), Multilayer Perceptron (MLP), and Decision Tree, Gradient Boost (GB), XGBoost (GB), Adaboost (GB), CATBOOST (GB), and LightGBM (LGBM). There is a wide range of precision amongst the models used here. A technology that can precisely predict diabetes is shown in this research. The findings of this research indicate that one of the Data mining models, random forest (RF), and the ANN Model from the category of neural network models have superior accuracy in making diabetes forecasts.