Evaluating the Performance of Machine Learning Algorithms for Diagnosing Diabetes in Individuals

Application of machine learning algorithms for the diagnosis of diabetes have become a trending research area, as effort to improve current techniques and methods used by health care institutions to determine the occurrence of diabetes in individuals is now given more attention than before. This study attempts to evaluate the performance of five (5) machine learning models on diabetic dataset using Python to predict the incidence of diabetes. Pima Indian diabetes dataset from UCI machine learning repository was used for the study. To ensure quality evaluation of the algorithms, a second dataset provided by Dr. John Schorling of the department of Medicine, University of Virginia was used for double evaluation. Result shows that Naive Bayes algorithm performs better when used for the prediction of diabetes.