Comparative Analysis of ANN, XGBoost, and GridSearchCV-Tuned FNN for Diabetes Prediction

Diabetes mellitus has become one of the major health challenges globally, and its detection at an early stage is necessary for efforts directed at preventing its progression and reducing complications. Hence, the study will discuss advanced deep learning models applied to enhance the early detection of risk and its management. This paper proposes a model for feedforward neural networks with tuning of hyperparameters using GridSearchCV, artificial neural networks, and eXtreme Gradient Boosting for classification and diabetes prediction. This dataset has been procured from a reputed medical source, with preprocessing done to ensure the robustness of model performance through normalization. GridSearchCV finds the hyperparameters that reduce overfitting and increase accuracy in the respective system by optimizing the FNN model. Comparative analysis between ANN and XGBoost proves that the FNN model with its hyperparameters demonstrates better predictive accuracy and precision with recall. The research epitomizes further the potentials of deep learning models and, more so, the FNN model with its optimal hyperparameters in enhancing early detection of diabetes and risk management, hence opening avenues for more individualized and more effective interventions in health care.