We developed forecasting models based on the advanced statistical (stochastic) and soft computing (computational intelligence) techniques for predicting high frequency data sets. Firstly, we used the standard statistical tools such as the autocorrelation/partial autocorrelation function, clustering, etc. to identify the chunks of information that are deemed essential for knowledge representation. Afterwards, (1) we proposed statistical methods to identify the relationship between the information granules; (2) based on the platform of granular or soft computing (G or SC) we developed and formal expressed the underlying mechanisms (models) that generate the observed data and, in turn, to forecast future values of the investigated process in managerial decision-making. The proposed intelligent approach is applied to the time series of USD/EUR exchange rates. We found that it is possible to achieve significant risk reduction in managerial decision-making by applying intelligent forecasting models based on the latest information technologies. We show that statistical GARCH-class models can identify the presence of the leverage effect and to react to the good and bad news. In a comparative study is shown, that both presented modeling approaches are able to model and predict high frequency data with reasonable accuracy, but the neural network approach is more effective and accurate.