Comparative Analysis: Personalized Random Forest Algorithm versus Gaussian Naive Bayes Algorithm for Diabetes Mellitus Classification in Humans

Abstract

In this research, a new Random Forest algorithm is evaluated against a Gaussian Naive Bayes algorithm to enhance the accuracy and precision of diabetes mellitus classification in individuals. The study employs the Pima Indian Diabetes Dataset sourced from Kaggle, a machine learning repository, featuring eight attributes used as input variables for the classifiers. Each classifier is tested with 20 samples per group. An assumed G power of 0.8 guides the study. The results indicate that the novel Random Forest algorithm achieved 91% accuracy and 85% precision, while the Naive Bayes method reached 78% accuracy and 84% precision. A significance level of p < 0.05 is employed. Notably, this investigation underscores the superiority of the Novel Random Forest method over the Naive Bayes algorithm for the accurate categorization of diabetes mellitus within the dataset.