Data stream has gradually become the hot topics of data processing circle. But the majority of the research work is on query of the data stream. Study on prediction of data stream is relatively little. Because of the limitations of traditional prediction methods, they can not meet the needs for real-time demand of data stream processing. Therefore, an efficient method of prediction of time series data stream using wavelet transform and LSSVM is presented. To ensure the prediction accuracy, wavelet transform is used to decompose the data stream into several layers, and then LSSVM is used to predict each layer. The wavelet transform incremental algorithm of improved SSWAT is presented, and incremental algorithm of LSSVM is combined to meet the need of time. The prediction of the power load and rotor angle data in the transient stability analysis of power system proves the effectiveness of the proposed method in prediction accuracy and speed, as well as the feasibility in power system application. |