Zeki Sistemler Teori ve Uygulamaları Dergisi

Yazarlar: Gürkan BİLGİN

Konular:Bilgisayar Bilimleri, Yapay Zeka, Mühendislik, Elektrik ve Elektronik

DOI:10.38016/jista.877292

Anahtar Kelimeler:Diabetes risk,Early stage,Support Vector Machine,Ensembles Learning,Algorithm,Decision Trees

Özet: Diabetes significantly affecting the quality of human life, the world and the incidence of a disease in Turkey is increasingly important. In particular, it causes damage to the nervous system, kidney, heart, eyes, limbs and blood vessels and can cause significant losses. For this reason, early diagnosis and follow-up is of great importance in order to prevent diabetes or to minimize the damage it will cause. Classification techniques obtained by machine learning algorithms have been accepted as important by researchers for the risk prediction model of the disease. In the study, a database created with information from 520 subjects was used to estimate the probability of developing diabetes. In the study, Multilayer Perceptron Artificial Neural Networks (MLPNN), Support Vector Machines (SVM), Decision Trees (DT), Ensemble Learning Algorithms (ELA), Linear Discriminant Analysis (LDA), k-NN Methods were used as machine learning methods. Among these methods, k-NN algorithm provided the highest accuracy and 99,81% accuracy was achieved with this algorithm. A diabetes early diagnosis kit was developed by including the algorithm providing the highest accuracy value into a computer user interface developed within the scope of the study

---