Classification of teletraffic service devices by κ-NN, ANFIS and ANN classificators
Ivelina Stefanova Balabanova, Georgi Georgiev, Pencho K. Penchev, Stela Savova Kostadinova, Rozalina Dimova · 2016
In this paper various types of classifiers for quantitatively identify teletraffic service devices are proposed. The classification method “K - Nearest Neighbors With Defined Cityblock Metric Distance At Three Nearest Neighbors” is selected. A classifier structure is synthesized based on Adaptive Neuro-Fuzzy Interface Systems (ANFIS) in hybrid learning algorithm and Gaussian type membership function of the input variables. Results are obtained for variation of mean square error and classification accuracy in a variation of neurons in the hidden layers of artificial neural networks with different number of output neurons and method of encoding target classes. A network structure with better performance is selected based on the parameters values of linear regression and correlation.