Application of Machine Learning Algorithms for Early Prediction of Diabetes using Lifestyle and Physiological Data

Abstract

Type 2 Diabetes Mellitus (T2DM) is a major global health concern, emphasizing the need for early detection to improve patient outcomes. This paper conducts a comparative study of four supervised machine learning classifiers, Two-Class Boosted Decision Tree, Decision Forest, Logistic Regression, and Neural Network, leveraging the Pima Indian Diabetes dataset within Azure ML Studio, a low-code cloud-based platform. Model performance was rigorously assessed using accuracy, precision, recall, F1 score, and AUC-ROC. Among the models evaluated, the Boosted Decision Tree achieved the most balanced performance with an accuracy of 0.774, recall of 0.663, and F1 score of 0.671. In contrast, Logistic Regression and Neural Networks demonstrated higher precision but lower recall, underscoring important trade-offs for clinical application. This study highlights the practical potential of low-code machine learning platforms in accelerating healthcare analytics. However, limitations include the specific demographic of the Pima dataset, lack of model interpretability methods, and no assessment of fairness. Future research will focus on model generalizability, integration of explainable AI techniques, and fairness evaluation. Overall, the results support the potential of accessible and responsible AI solutions to advance preventive healthcare.

Ähnliche Arbeiten