Improving Classification Performance for Diabetes with Linear Discriminant Analysis and Genetic Algorithm

Abstract

In the modern-day, Diabetic disease is one of the most chronic and appalling diseases humanity faces. There are 463 million people had Diabetes worldwide, and it caused approximately 4.2 million deaths, according to the International Diabetes Federation (IDF) Diabetes Atlas Ninth edition 2019. Therefore diabetic patients need state-of-the-art healthcare against such diseases and propose early prediction to help decrease the risks related to such diseases. In this context, this research, a diabetes diagnosis system, has proposed to analyze two different diabetes datasets, namely PIMA Indian Diabetes and data of Dr. John Schorling. Linear Discriminant Analysis (LDA) and Genetic algorithm (GA) methods used for feature selection and four techniques implemented to evaluate the classification are Bagging algorithm, Random forest, Logistic Model Tree (LMT), and JRip algorithm. The results have shown that a random forest classifier using LDA and GA obtained better accuracy (90.89%) in DatasetI. At the same time, DatasetII is better than GA in Random forest, random forest-LDA, JRip-LDA classifiers (91.44%).