This paper proposed a time series data stream prediction method.By empirical mode decomposition, decomposed the original dataset from re-writable linked window into finite intrinsic mode functions (IMFs) and a residue.Then forecasted each component by maximal Lyapunov exponent method.Finally combined the predicted results of all components into the final value.Experiments with power load show that the combined model has higher precision of prediction and better adaptability compared with other single time series prediction models.%提出一种通用的时间序列数据流预测方法,算法首先通过经验模式分解方法将从链式重写窗口取得的数据集分解有限具有特征振荡周期的固有模态函数分量和一个代表原始序列平均趋势的余量;然后对于各个分量分别建立最大Lyapunov指数预测模型进行预测;最后将各分量的预测值组合获得最终预测值.通过电力负荷的预测实验表明,与单一的时间序列数据流预测模型相比,该模型具有较高的预测精度和很好的模型适应性.
展开▼
