DefinePK

Abstract

Diabetes represents a significant global health challenge, particularly as the prevalence of individuals at risk continues to rise. Diabetes is classified as a chronic disease; diabetes is responsible for a substantial number of fatalities annually. Early prediction of diabetes is essential for halting its progression and mitigating the risk of severe associated complications, including cardiovascular disease and renal damage. This research proposes an innovative Deep Learning (DL) clinical decision support system designed to optimize diabetes prediction accuracy using machine learning. The proposed DL methodology combines a stacking pattern based on learning with various DL architectures, specifically ANN, LSTM networks, and CNN, to form a Combined Learning Network Pattern (CLNet). To enhance diabetes prediction capabilities, the DL framework employs a pattern that integrates meta-level patterns. The novel DL patterns are trained to utilize three distinct diabetes information sets. Pertinent variables are obtained from the information set utilizing the proposed methodology. Key evaluation metrics such as accuracy, precision, recall, specificity, F1-score, MCC, and ROC/AUC are employed to assess the effectiveness of the proposed CLNet patterns. When applied to the information sets, the Combined Learning Network Pattern (CLNet) exhibited improved performance compared to the other proposed CLNet pattern, obtaining accurate results rates of 99.5%, 98.8%, and 98.4%, respectively. The analysis reveals that the proposed CLNet patterns exhibit enhanced performance in diabetes prediction compared to previous studies.