A Modified Multivariate Bayesian Logistic Model with Application to Health Datasets

Abstract

The application of Bayesian strategies for binary logistic estimation is demonstrated in this article. A modified method of the Bayesian logistic model using the Metropolis-Hasting algorithm is derived and applied to three simulation data sets. We compared the new model with existing classification methods: support vector machine, artificial neural network and regular logistic model. The modified model was used to classify the heart disease dataset. The data came from a database intended for UCI Data Science (https://www.kaggle.com). The clarification accuracy and the time required are checked and compared with other standard methods. It has been shown that the presented model has the best accuracy and efficiency compared to the different classification methods. All calculations were performed with the program R version 4.2.2.