Evaluation and Measuring Classifiers of Diabetes Diseases

Abstract

Classification plays tremendous role in data mining process, especially for huge amount of data and it is suitable for predict new knowledge and discover patterns. This process can work with different types of data whether it was nominal or continuous. In this paper classification will be performs on diseases diagnoses by choosing to work with (k-nearest neighborhood algorithm KNN) measure and evaluate the method with (Artificial Neural Network ANN). These two classification methods have been chosen to classify (Pima-Indian-Diabetes PID) using spiral spinning technique. Classification done by taking 1 to 50 values of (K) in KNN versus 1 to 50 values of hidden layers for ANN in single iteration checking the accuracy as measuring to evaluate performance. T-test used to validate choosing two different factors (K in KNN and number of hidden layers in ANN), t-test results shows that the method is extremely statically significant. After performing classification by changing architecture, ANN proves better results than KNN in this disease classification.