An approach is presented for predicting both the amplitude and phase responses of a channel. The technique can be used to predict any type of channel, and a priori knowledge about the channel is not necessary. But any available information about the type of channel can easily be incorporated within the technique, so that it leads to better prediction. The stability problem, which is the main drawback in DFE (decision feedback equalizers) is overcome by data assisted equalization, which gives reliable estimates of the channel even at low SNR (signal-to-noise ratio). Convergence time is reduced. Finally, a method for implementing the equalizer is presented and a discussion is outlined pertinent to a suitable adaptation algorithm.