Possibility Prediction Of Diabetes Mellitus At Early Stage Via Stacked Ensemble Deep Neural Network

Abstract

Diabetes Mellitus is a chronic metabolic disease caused by the deficiency of insülin action or secretion, or both, one of the hormones that balance the blood glocose level. It is one of the health problems that negatively affect people's quality of life. If diabetes is not detected in the early stages, it can cause serious complications such as heart and renal diseases, retinopathy, stroke, digestive disorders, and amputation. Because of the presence of a long asymptomatic period, early detection of diabetes is not realised usually. For this reason, around 50% of diabetic patients are not received a treatment due to undiagnosed at early stages. This situation results other diseases mentioned above, which diabetes causes. On the other hand, ensemble learning is a machine learning model in which multiple models are trained to solve the same problem and combined to achieve better results. Deep neural networks are one of the machine learning algorithms and they are the multi-layered state of artificial neural networks developed inspired by the information processing method of the human brain. In this study, a stacked ensemble-based deep neural network approach is proposed for diabetes possibility assessment in the early stages. The proposed approach was tested on a dataset of 520 patients. As a result, the proposed method achieved the highest success rate with 99.36% accuracy and 99.19% AUC, although the test percentage was kept higher than the prediction studies conducted on the same dataset.