| Abstrakt: |
Hyperglycemia or high blood glucose levels, is a severe symptom of diabetes mellitus. Machine learning approaches that are implemented correctly can help in the early detection of diabetes. Particle swarm optimization (PSO) features are selected from Diabetic patients' dataset which are classified through four classifiers namely, Logistic Regression, Expectation Maximization (EM), Firefly and Firefly with EM classifier. The University of California, Pima Indians Diabetes patients databases also known as the UCI machine learning online database, provided the diabetic patients dataset is analyzed in this research. The performance of the classifiers were examined and compared based on parameters like, Sensitivity, Specificity, Accuracy, Performance Index (PI) and Mean Square Error (MSE). Results show that the Logistic Regression classifier for diabetic case exceeds the other three classifiers in terms of accuracy (93.36%), sensitivity (86.72%), specificity (100%), and PI (84.66%). The EM classifier for normal case exceeds the other three classifiers in terms of accuracy (98.96%), sensitivity (97.92%), specificity (100%), and PI (97.87%). [ABSTRACT FROM AUTHOR] |
