Author(s)

Philip de Melo, Marie St. Rose

Department of Nursing and Allied Health, Norfolk State University, Norfolk, VA, USA.

The recent surge in demand for timely and accurate health information has highlighted the need for more advanced data analysis tools. To reduce the incidence of preventable medical errors, sophisticated IT-driven classification and prediction algorithms are essential. However, extracting meaningful insights from complex biomedical data remains a significant challenge in healthcare transformation. Modern biomedical and health research generates diverse data types, including electronic health records (EHRs), medical imaging, sensor data, and telemedicine inputs, which are often complex, heterogeneous, poorly annotated, and largely unstructured. Traditional statistical learning and data mining methods require extensive preprocessing before developing predictive or clustering models. This process becomes even more challenging when dealing with intricate datasets and limited domain-specific knowledge. Recent advancements in deep learning offer promising end-to-end models capable of handling such complexity. However, these models do not consistently achieve the high levels of accuracy required by healthcare professionals. In this study, we introduce a novel Deep Learning Algorithm combined with a generative AI designed to improve classification accuracy in clinical applications significantly. The algorithm is tailored for seamless integration into hospital workflows and electronic health record systems—an area that is the central focus of our ongoing research. The proposed method combines real-world clinical data with synthetic data generated by Principal Model Generative AI. This approach increased classification accuracy in our experiments from 76% to 95% - 98%.

Diabetes, Informatics, PIMA Data Set, Deep Learning, Improved Accuracy, PM Generative AI

Melo, P. and Rose, M. (2025) Accurate Classification of Diabetes via PM Generative AI. Advances in Bioscience and Biotechnology, 16, 379-409. doi: 10.4236/abb.2025.169025.