Wykorzystanie sztucznych sieci neuronowych przez przedsiębiorstwa eksportujące oraz importujące

The paper raises issues of the use of artificial intelligence in the enterprise. The focus was on the possibility of using artificial neural networks to accurately predict the behavior of the time series relevant to the economic activities based on the export and import. In particular, the paper describes the practical possibilities for time series forecasting such as foreign exchange rates. Researches focused on predicting of slope of linear regression, to determinate the direction of exchange rate changes. Artificial neural networks, tested during researches, included two types of models. First one was a simple neural network model, containing only a one network. Second one was a more complex model containing at least a few networks. These networks were used for predicting a part of output variable. To obtained mentioned parts during researches was used multiresolution analysis based on discrete wavelet trans- form. During researches a lot of versions of multiresolution analysis were tested. Finally, as the best one, was chosen the discrete wavelet transform based on the biorthogonal 6/8 wavelet. The paper describes also a type of model input variables, considering a frequency of their changes. It shows advantages and disadvantages of macroeconomic data and technical analysis. The article describes main and the most useful types of moving averages, such as simple moving average, exponential moving average, weighted moving average and VI- DYA (Variable Index Dynamic Average). The paper mentions other type of input variable, especially such indicators as RSI and MACD and their modifications. The final evaluation of the models was carried out based on a simple trading system. Thus was confirmed the useful- ness of the results in practical applications. During the analysis of the obtained results, was used the method of sliding window.