Performance evaluation of machine learning classification techniques for Diabetes disease

Diabetes is a noncontagious disease where Diabetes of type two Mellitus is among the top five leading the cause of global death. Not knowing the status of patients leads to complications such as kidney neuropathy and retinopathy, eventually lead to death. Knowing the patient’s stand using machine learning techniques can assist in early treatment will be useful in lowering the burdens mentioned above caused by Diabetes. In this work, researchers focused on evaluating the patient’s status of Diabetes. In this study, the Cross-Industry Standard Process for Data mining (CRISP-DM) used as a research methodology of research. Where Support Vector Machine, Decision Tree, Naive Bayes used as a classification technique, the study aims to predict the patient status for optimizing the complication caused by Diabetes. The data set used for the model was retrieved from the Pima Indian diabetic database Diabetes Database (PIDD), which is obtained from the UCI machine learning database with 768 records in total. KNN algorithm can be made best with an accuracy of 76% for the condensed dataset with the nine attributes as identified from the comparison of the result of different models.