This article provides the comparison of libraries neural networks. Based on this analysis was determined to develop a neural network for classification of spectra based on Encog library, because it implemented many components and gives the best result with a small number of items for training. Showed the architecture of neural networks for data classification of spectral analysis.
The moving average method with the 4 samples window width is used to raise the weekly forecast of the US dollar exchange rate accuracy. The non-iterative artificial neural network with the radial basis functions is used for. In the end we got the forecast error less than 1%
The IAS "Forecast" is developed for forecasting the electricity consumption in the original production conditions at PJSC "Lvivoblenergo." The statistical and neural network methods are used for the input data verification; is enhanced the space dimensions extending methods for the incoming data to use them with the ANN with non-iterative training