Enhancement performance of the Naïve Bayes method using AdaBoost for classification of diabetes mellitus dataset type II

Abstract

In using technology, especially in health sciences, machine learning modeling can make it easier to predict disease treatment. Naïve Bayes optimization with AdaBoost is needed because even though Naïve Bayes has the advantage of minimal parameters, its accuracy is susceptible to too many features. AdaBoost is used to overcome sensitivity to an excessive number of features and optimize its ability to handle complex datasets. This research aims to analyze the classification results of the Naïve Bayes method with the help of the AdaBoost method. This data comes from Community Health Centers I, II, and III Mengwi District, Bali Province patient medical records. The classification process uses the Naïve Bayes method and Naïve Bayes with AdaBoost, which is then evaluated using a confusion matrix. Two scenarios were used in testing: Naïve Bayes and AdaBoost-based Naïve Bayes. The algorithm is implemented on the dataset and tested directly using cross-validation. The evaluation results show that the Naïve Bayes method experienced an increase in accuracy of 5.92% at 5-fold and 5.93% at 10-fold on a dataset with 890 data. The addition of the AdaBoost method to diabetes classification has been proven to improve the accuracy performance of the Naïve Bayes method.