부산대학교 도서관

A host server load state prediction algorithm based on nonlinear time series analysis and vector space reconstruction is proposed. The load state of the host server is analyzed discretized by the phase randomization method and decomposed into a statistic containing several nonlinear components. The delay is obtained by using autocorrelation function method in vector space reconstruction. A vector space reconstruction of load state based on mutual information minimum embedding dimension is proposed. By extracting the high-order spectral characteristics of the load in the high-dimensional vector space, the network service is accurately predicted. Simulation experiments show that the proposed method can adapt to the nonlinear characteristics of traffic flow well, and can track the change of network traffic condition better, and its prediction error is lower than that of the traditional method.