the channel prediction is for compensating the feedback delay. With channel prediction, the feedback parameters are calculated from a predicted channel, such that the calculated feedback parameters fit the channel at the actual time of transmission.
You find a description of the implemented methods in my master thesis
there they are described as channel interpolation, however, methods are used in terms of extrapolation for channel prediction
To see the impact of channel prediction you need to set the following parameters: downlink delay >0, user speed >0, activate feedback, activate channel prediction, use correlated fast fading channel model