A Machine Learning Framework for Diabetes Detection Using Hoeffding Tree

This research aims to classify diabetes patients using machine learning algorithms, with a particular focus on the Hoeffding Tree model. The study utilizes a dataset comprising 256,380 samples, categorized based on ten key health and demographic attributes: (1) history of stroke, (2) daily vegetable consumption, (3) heavy alcohol consumption, (4) health insurance coverage, (5) inability to afford medical consultation in the past 12 months, (6) self-rated general health status (ranging from excellent to poor), (7) difficulty in walking or climbing stairs, (8) gender (male, female), (9) educational attainment (from never attended school to college graduate), and (10) classification type. The objective is to develop a standardized and efficient diabetes classification model that reduces diagnostic time and optimizes resource allocation. Before selecting the Hoeffding Tree as the primary model, the research team conducted a comprehensive evaluation of 27 machine learning algorithms. The Hoeffding Tree demonstrated the highest classification performance, achieving an accuracy of 85.9528%. The findings highlight the model's effectiveness in diabetes classification and its potential for practical implementation in healthcare decision-making.