The Holographic and Perceptron Neuron Networks Joint Application for the Dynamic Systems Behavior Forecast

Abstract

The method of prognosis of dynamic system behavior along its temporal series is suggested. This method is based on the joint application of two types of artificial neural networks: holographic and perceptron. The holographic network plays the role of associated memory and perceptron one—the forecasting approximator. The method was tested on the task of forecasting electrical energy consumption. The holographis network selected similar intervals in the preceding temporal series. The perceptron was then trained on the selected data, using it as input values and the intervals immediately after them as output values. A trained perceptron may be used for the prediction. With the real data from the regional electrical company, the average relative error of the forecast was about 2%.