Forecasting via Wavelet Denoising - the Random Signal Case
In the paper we evaluate the usability of certain wavelet-based methods of signal estimation for forecasting economic time series. We concentrate on extracting stochastic signals embedded in white noise with the help of wavelet scaling based on the non-decimated version of the discrete wavelet transform. The methods used here can be thought of as a type of smoothing, with weights depending on the frequency content of the examined processes. Both our simulation study and empirical examination based on time series from the M3-JIF Competition database show that the suggested forecasting procedures may be useful in economic applications