Big data analytics makes predicting human behavior possible, but it is unclear how to exploit the predictable information for improving performance of wireless networks. In this paper, we investigate the potential of predictive resource allocation in supporting high throughput by exploiting excess resources. To this end, we assume that the requests and trajectories of mobile users and the average resource usage status of base stations can be predicted within a window. To fully use resources within the prediction window and reserve resources for the unpredictable traffic arrived after the window, we optimize a resource allocation plan to minimize the maximal transmission completion time. To assist the base stations for user scheduling, we introduce a method to make a transmission plan. These two plans determine where, when and what to transmit to the users with how much resources. Simulation results show that the predictive resource allocation can provide substantial gain over non-predictive strategy in terms of both network throughput and user experience.