Early Diabetic Diagnosis Patient Prediction based on Multilayer Perceptron DL Model

Abstract

Diabetes is one of the most rapidly increasing chronic diseases worldwide and early detection is essential for effective treatment and prevention of severe health complications. Accurate prediction of diabetic patients at an early stage can assist healthcare professionals in making better clinical decisions. In recent years, ML and DL techniques have gained significant attention for medical data analysis and disease prediction. This study presents an intelligent prediction model for early diabetic diagnosis using a MLP based DL approach. The proposed model utilizes medical attributes such as blood pressure, age, insulin level, body mass index (BMI), glucose level, and other health factors to forecast the risk of diabetes. To enhance the quality of the input characteristics, the dataset is first pre-processed using normalization and data cleaning methods. A Multilayer Perceptron neural network with several dense layers, batch normalization, and dropout layers is used to train the data after preprocessing to improve model generalization and lessen overfitting. The experimental findings show that the suggested MLP-based deep learning model reliably predicts outcomes and successfully learns intricate patterns from medical data. According to the simulation results, the suggested model can predict diabetic patients in their early stages with an accuracy of 87.68%. Therefore, the developed system can serve as a supportive decision-making tool for healthcare professionals to identify high-risk diabetic patients at an early stage and improve preventive healthcare management.