Özet:

Abstract Hyperglycemia which constitutes a considerable imminence to human health.  Diabetes may lead to an anomalous rise in glucose levels. Preliminary detection of diabetes reduces the risk of fatality and agony.  In our country around 30 million peoples are recognized with this fatal disease. It is tremendously complicated to develop a virtuous and precise diabetes forecasting. The ICMR with diabetic people have taken inventiveness and emerged with various solution but regrettably they endured like leftovers. Clustering is an important technique for the prediction of diabetes. In machine learning the clustering technique contingent on unsupervised learning and classification techniques contingent on supervised learning. In this research work, the factor analysis concept has been solicited to genesis of total effect on the PIMA Indian Diabetic Dataset and designate the prime factors that repercussion on it. K-Means algorithm conviction has been on the total effect data to acquire the cluster in superlative mode and for the quantification of distance the Euclidean distance function has been used. The numbers of clusters have been pronounced on the base of output of the dataset and it causes formation of knowledge based. To predict diabetics various machine learning accession have been solicited on cluster-based dataset. K-Means clustering algorithm used for early diabetic identification containing the data of 165 diabetic patients. The maximum precision, recall and F1-score1.00 obtained by K-Means and accuracy obtained by logistic regression 0.7662, decision tree 0.7269, SVM 0.7835 and random forest 0.7922 respectively. All anticipated outcomes are displayed in a comparison table and pointed out the aspect of research.

Anahtar Kelimeler: