—Base station side precoding takes the advantage of channel reciprocal properties in Long-Term Evolution Time Division Duplex (LTE-TDD) systems to achieve high data rates for multiple users. However, for proper precoder design, the base station needs accurate up-to-date channel information. In this paper, the forward-backward autoregressive (AR) process is employed to predict the downlink channel of LTE-TDD systems. The proposed integrated block diagonalization (BD) precoding and channel predicting method can improve the sum system capacity in a time-variant Rayleigh fading channel environment that is based on a Jakes model. The performance is compared with the conventional BD precoding method where channel prediction is not present. The proposed approach is supported by computer simulation results which indicate that there are significant system capacity improvements under certain conditions.