AutoML vs Classical Machine Learning: A Comparative Analysis Using Diabetes Prediction as a Case Study

Abstract

Automated machine learning (AutoML) is transforming how predictive models are developed, yet its real-world performance compared to expert-engineered pipelines remains underexplored. This study systematically evaluates AutoGluon AutoML against classical machine learning using diabetes prediction as a representative clinical case. Across binary (100,000 Kaggle samples) and 3-class ADA 2025 experiments, AutoML matched expert pipelines in predictive accuracy (AUC0.9790 vs 0.9791) while reducing development effort by 93%, requiring minimal technical expertise. Statistical analysis confirms no significant performance difference. Beyond efficiency gains, AutoML’s automated pipeline construction demonstrates potential for integration into emerging Agentic AI systems. These findings position AutoML as a practical, scalable, and democratizing tool for clinical decision support and broader data science applications, highlighting both immediate utility and future avenues for autonomous, high-impact predictive modeling.

Mohamad Abou Ali

Postdoctoral Researcher

I develop generalizable deep learning methods for biomedical imaging, advancing diagnostic accuracy, data augmentation, and intelligent healthcare systems.

Fadi Dornaika

Ikerbasque Research Professor

Ikerbasque Research Professor with expertise in computer vision, machine learning, and pattern recognition.

Ignacio Arganda-Carreras

Ikerbasque Research Associate Professor

My research interests include image processing, computer vision, and deep learning for biomedical applications.