Abstract:
The paper aims to verify the ability of artificial neural networks to model and predict time series with seasonal and trend pattern. In this study the effectiveness of data preprocessing and time series analysis is examined, especially deseasonalization and detrending as a basis for further neural network modelling and forecasting. In this paper it is proved that using deseasonalization as data preprocessing method, the best neural network performance is reached with respect to smallest Mean Squared Error showing the difference between outputs and targets. In general the research shows that prior data preprocessing enhances preciseness of further neural network prediction.
