OPTUNA—Driven Soft Computing Approach for Early Diagnosis of Diabetes Mellitus Using ANN

Diabetes is a severe disease characterized by elevated blood glucose levels. Early identification and prediction of diabetes are crucial for effective management. Considering that diabetes affected approximately 422 million people globally in 2018, the World Health Organisation (WHO) has reported, that the significance of accurate classification and prediction methods cannot be overstated. This research contributes in enhancing the overall accuracy of diabetes prediction. The proposed work introduces an enhanced method for diabetes classification and prediction by integrating artificial neural networks (ANNs) into the OPTUNA hyperparameter optimization framework. Soft computing techniques, including Recursive feature elimination with cross-validation (RFECV) and principal component analysis (PCA), are employed to improve performance and handle uncertain and imprecise data. Outcomes of experiments on the Pima Indian Diabetes dataset demonstrate superior accuracy compared to conventional methods. The proposed approach offers a powerful decision support system for healthcare practitioners, aiding in early diagnosis and effective management of diabetes. The proposed methodology achieved high accuracy rates of 98.78% for the trained data and 98.44% for the test data, demonstrating its effectiveness by combining the strengths of ANNs and soft computing.