Automated Machine Learning And Deep Learning Techniques For The Classification Of Diabetes Mellitus: A Model Development Study
##plugins.themes.bootstrap3.article.sidebar##
Download : 46 times
##plugins.themes.bootstrap3.article.main##
Prof. Sanmati Kumar Jain
Dr. Sheetal Bawane
Abstract
A dangerous disease in humans, diabetes mellitus is caused by elevated glucose levels. Untreated diabetes may cause a host of further serious health complications. This research aims to predict the occurrence of diabetes by analysing several human bodily characteristics. For the purpose of diabetes mellitus risk prediction, an ensemble method called En-RfRsK is suggested. This voting classifier combines three machine learning techniques: RF, R-SVM, and KNN. RF makes use of the results obtained from a multitude of models or trees that are not always evenly distributed. Using a function whose value varies as the distance from the origin increases, R-SVM is able to make predictions. By analysing diabetic data, KNN is able to understand the non-linear decision limits. This new method makes use of all the best features of various ML approaches. Since ensemble methods outperform single classifiers in terms of accuracy and flexibility, the suggested method is an amalgamation of preexisting ML techniques. Because of its superior prediction powers and accuracy, it provides the most suitable answers. Using the PIMA diabetes dataset, experiments were conducted. Experiments clearly show that the suggested method beats both the current baseline classifiers and the most cutting-edge ML diabetes mellitus prediction systems. An accuracy of 91% was achieved using the En-RfRsK method.
##plugins.themes.bootstrap3.article.details##
This work is licensed under a Creative Commons Attribution 4.0 International License.