A user request prediction method based on data mining approaches has been proposed in this paper. By analyzing a large amount of learning history data accumulated over a long period of time, the presented solution enables the coming request to be predicted in advance, and preprocessed beforehand. It takes advantage of the integration of data mining technology and parallel computing, which maximizes the server's ability of parallelization from the knowledge acquired from the experiential data. The proposed protocol has been applied to a practical e-learning system, and the experiment result shows that the average response time with the proposed scheme is much shorter than the one without request prediction