Examination of Diabetes Mellitus for Early Forecast Using Decision Tree Classifier and an Innovative Dependent Feature Vector Based Naive Bayes Classifier

Abstract

The proposed study aims to evaluate the accuracy and precision in earlier diabetes mellitus detection using Decision Tree and Naive Bayes classification algorithm. Materials and methods: Naive Bayes classifier is applied on a Pima Indian diabetes dataset that consist of 769 records. A machine learning technique for earlier prediction of diabetes disease which compares Decision Tree and Naive Bayes classification algorithms has been proposed and developed. The sample size was measured as 27 per group. The accuracy and precision of the classifiers was evaluated and recorded. Results: The accuracy was maximum in predicting diabetes using Naive Bayesian classifier (76.46%) with minimum mean error when compared with Decision Tree Classifier (70.09%). There is a significant difference of 0.006 between the groups. Hence, Naive Bayes appears to be better than Decision Tree Classifier. Conclusion: The study proves that Naive Bayesian Classifier exhibits better accuracy than Decision Tree Classifier in predicting diabetes.