Comparison of classical regression and neural networks in exchange rate forecasting

This diploma thesis deals with the exchange rate forecast using the method of artificial intelligence and classical regression. The theoretical part consists of four chapters. The first chapter describes all important information concerning exchange rates – exchange rate, exchange rate regimes, factors influencing the exchange rate from the point of view of the time horizon, determination of possible exchange rate determinants for the following forecast. The second chapter is devoted to time controllers and various regression models for your prognosis. The third chapter describes the methods of machine learning and the specific method of neural networks, which is further used in their own work. The last theoretical chapter tells about the methodology for CRISP-DM data mining in accordance with the practical part. In the practical part, the author forecasts the EUR / USD exchange rate in conjunction with the theoretical part. Exchange rate prediction is given according to the CRISP-DM methodology, which includes data preprocessing, model construction and their validation. Regression models and neural network models are applied to the exchange rate forecast. In conclusion, the author compares the predictive capabilities of both models on test data.