A Comparative Evaluation of a Machine Learning-Based Probability Algorithm for Early Diagnosis of Diabetes

Abstract

This study evaluates a novel machine learning-based probability algorithm for predicting 
diabetes risk based on clinical and laboratory parameters. The model is derived from a dataset of 500 
patients from a national health database and tested across three independent cohorts from the USA, 
India, and Germany. The performance of the proposed model is compared with a Bayesian probability 
algorithm derived from existing medical literature. Results indicate that the machine learning model 
provides superior accuracy and reliability, particularly in populations with intermediate disease 
prevalence.