PERBANDINGAN ALGORITMA CART DAN NAÏVE BAYESIAN PADA KASUS DIAGNOSIS PENYAKIT DIABETES

Abstract

Diabetes mellitus is a disease that threatens serious health, can cause death and the World Health Organization (WHO) estimates that every 10 seconds there is one diabetes patient who dies of this disease. This makes researchers and practitioners focus their attention on detecting / diagnosing diabetes mellitus and preventing it because this disease can cause complications. The method used in this research is problem identification, data collection, pre-processing stage, classification method, validation and evaluation and conclusion drawing. The algorithm used in this study is CART and Naïve Bayes by using a dataset taken from the UCI Indian Pima database repository which consists of clinical data of patients who detected positive and negative diabetes mellitus. The validation and evaluation methods used are 10-cross validation and confusion. Matrix for precision, recall and F-Measure. The results of calculations that have been done, the results of the accuracy of the CART algorithm are 76.9337% with precision 0.764%, recall 0.769%, and F-Measure 0.765%. While the diabetes dataset tested by the Naïve Bayes algorithm gets an accuracy value of 73.7569% with precision 0.732%, recall 0.738%, and F-Measure 0.734%. From these results it can be concluded that to diagnose diabetes mellitus it is recommended to use the CART algorithm.