Moving trend based filters design in frequency domain

An original approach to digital moving trend based filters (MTF) design, based on Bode plots analysis is proposed, aimed at seasonal time series decomposition and prediction. A number of polynomials of different range are discussed to be used in the MTF as the LS approximation formula. The Bode plots of the MTF are shown, and the best filter is selected. Results of a seasonal time series decomposition and prediction with the best MTF is presented and compared to the classical MTF calculations (involving the linear LS approximation).