Common cooperative relay selection used outdated Channel State Information (CSI) to select relay, and accurate channel prediction can provide precise CSI for relay selection. The existing channel prediction methods often cannot adapt to the rapidly time-varying channel, or with high complexity. Therefore a channel prediction method based on slope prediction was proposed, which was applied to relay selection for cooperative communication system. This channel prediction method combined first-order linear polynomial curve fitting with Finite Impulse Response (FIR) Wiener prediction, by using FIR Wiener predictor to predict the slope of channel first-order linear function, cut the prediction range into several small enough time slices, and implemented continuous channel prediction within these small segments of time. At the same time, the Channel-Prediction based Relay Selection (CPRS) method was proposed, and performance analysis was carried out. Simulation results show that in the comparison experiments with the coordination system of Outdated-Channel based Relay Selection (OCRS), CPRS can make the system Bit Error Ratio (BER) decline 13% to 63%, with significant performance improvement effect.