A key issue is a development, testing and comparison of different prediction methods.
The advanced methods using e.g. neural networks, minimum variance prediction, Kalman
filtering are compared with the classical ones, like naive prediction of different types,
smoothing or decomposition. The methods of a’priori estimation of prediction efficiency are
developed and compared with an ex post prediction efficiency obtained with the use of a
particular prediction algorithm. A set of new computer programs dedicated for testing the
different prediction methods has been designed and developed. The programs written Matlab
or C++ are user friendly and make possible testing the properties of the prediction methods in
dependence on the assumed parameters.