Predicting Diabetes Using Diabetes Datasets and Machine Learning Algorithms: Comparison and Analysis

Abstract

The performance of three popular machine learning algorithms to predict diabetes based upon using three diabetes datasets is presented. Two of the datasets are from the public domain and the third is composed from a research study group. The J48, Random Forest and Naïve Bayes machine learning algorithms were evaluated. Machine Learning (ML) is used to both analyse and make predictions from data that is simply too voluminous for humans to process. This is especially true with medical data where the use of machine learning and data analytics is still in its infancy. More specifically this research investigates the application of ML algorithms on the growing data from the healthcare industry on the global diabetes epidemic. The performance of ML algorithms to predict diabetes is lacking. This paper provides an analysis of the challenges of machine learning in this field and covers this gap in the research.