Mathematical methods for forecasting bank transaction data

We aim at comparing different methods for forecasting several types of credit card transaction data. These data can be devided into two groups – one of them is characterised by a distinctive structure (trend, seasonality) but with very short data bases, the other one is larger but high volatile. Evaluated methods include those from statistics and nonlinear dynamics as well as wavelet techniques. Beside the software developed for this purpose standard toolboxes for time series analysis have been applied.