Diabetes Prediction Based on Machine Learning Techniques: A Review

Abstract

Early and precise diabetes diagnosis is vital for efficient disease regulation and complication prevention. The emerging role of machine learning (ML) in healthcare has given impetus to researchers who have explored various predictive models to enhance diagnostic precision. This review critically examines tree-based methods, probabilistic and distance-based approaches and neural network techniques, evaluating their performance based on various metrics. Our analysis indicates that ensemble learning approaches, particularly random forests with gradient boosting, along with artificial neural networks, consistently outperform traditional models, exhibiting superior predictive capabilities. These outcomes emphasise the potential of ML in the betterment of diabetes detection by identifying complex patterns in patient data. Integrating advanced predictive algorithms into diabetes screening can enhance early detection and enable timely medical interventions. With machine learning models continuing to evolve, their application in medical diagnostics holds significant promise for bettering diabetes detection and assisting healthcare professionals in selecting the most effective predictive models for clinical use.

DOI: https://doi.org/10.24321/3051.4304.202602