ENHANCING EARLY DIABETES DETECTION A COMPARATIVE ANALYSIS OF MACHINE LEARNING AND DEEP LEARNING TECHNIQUES

Abstract

This research paper examines the use of machine Learning and Deep Learning technique in prediction of diabetes and early detection to control the spread of diabetes globally. The study uses a comprehensive methodology which includes data preprocessing, feature engineering, and traverses through implementation of various predictive models on the dataset utilizing PIMA Indian Diabetes dataset. Then, traditional machine learning algorithms (Logistic Regression, K-Nearest Neighbors, Support Vector Machines, and Random Forest) performance are compared with deep learning models (Feedforward Neural Networks and adapted Convolutional Neural Networks). The study evaluates the approaches to the problem on the basis of rigorous evaluation metrics and further through statistical analysis to find the most effective method in prediction of diabetes. The results provide one piece in a growing pool of knowledge on how AI can be applied to healthcare, with potential to enhance the early diagnosis and treatment of diabetes. Moreover, this research provides insights not only into the strengths and weakness of current predictive models for medical AI, but also indicates directions for future work in this field.