INTERNET TRAFFIC ZONE IDENTIFICATION BY BACKPROPAGATION AND PROBABILISTIC NEURAL NETWORKS

The article proposes an approach based on the concept of Artificial Intelligence for categorization urban areas Internet content by corporate customers. applicability different neural apparatus was analyzed as well three-layer Backpropagation Neural Networks (BPN) and four-layer Probabilistic (PNN) most suitable purpose study were selected. synthesis BPN architectures traffic identification carried out according to a number computing units in hidden layers with hyperbolic tangent sigmoid, log-sigmoid linear transfer functions. variations set specific criteria examined Accuracy, Mean-Squared Error, Mean Absolute Correlation coefficients, etc. selection PNNs against defined quality indicators stepwise increase spread indicator Kernel functions Radial-Basis (RB) structural layer analogy similar that applied BPNs. In research processes, high levels recognition established processing Incoming flows Packages Accuracy over 90.00%.