Porkhun O. Automatic classification of multidimensional objects using neural networks

The method for determining the number of clusters during the clusterization by the Kohonen neural network using a criterion of quality is offered. The criterion combines density and distance based measure and method of ideal point. The method for constructing features vector for classification presented by the set of heterogeneous parameters is offered. Using the feedforward neural network and the method for constructing features vector the automatic system of texts classification presented by the selected set of parameters is developed. Using the offered method for determining the number of clusters the automatic system for texts clustering is developed. The developed systems were applied for the decision of tasks of art texts attribution.