| Abstract [eng] |
The main purpose of this paper is to develop a model to predict future exchange rates using the LSTM algorithm of an artificial neural network, after reviewing previous related studies and evaluating their methods and results. Without introduction and counclusion the thesis consists of three main parts: theoretical aspects of foreigh markets and artificial neural networks, methodology, practical application of the developed exchange rates prediction model. In the first chapter of the thesis the most important aspects of the international exchange rates market presented and the the key factors influencing exchange rate changes indentified. It also neural networks main concepts, classification, network training analized and previous related studies and their using methods reviewed and investigated. In the methodology based on the research of artificial neural networks, a model that can predict the selected exchange rates in the futured created. Algorithm and parametres being used in this thesis was explained in details with their backend mathematical formulas in order to explain the detailed plan for further research. In the empirical part, using the Python programming language and an abundance of libraries, the currency rate prediction model created in methodology is used to perform experimental tests on historical data for selected exchange rates pairs. Using long-term short-term memory (LSTM) neural network algorithm, to perform exchange rate forecasting and statistical results to verify the obtained results programming is used. After checking the results obtained and the suitability of the model, the exchange rates of the currency pairs for the selected periods are forecasted. In conclusion the results of the thesis presented, main concepts of the research summarized and recommendation for future improvements provided. |
