-
摘要: 直接多步预测不依靠单步预测的结果而直接使用测量数据, 效果理想, 但往往要求模型能够学习多种不同的目标函数. 本文建立的直接多步预测混合模型, 使用模式分解方法把原始时间序列分解成不同尺度的基本模式分量, 再经混沌分析和神经网络进行组合预测, 减小了各步预测模型之间的差别, 提高了模型对多种目标函数的学习能力, 有效提高了预测精度. 最后, 通过基准时间序列验证了本模型的优越性.Abstract: The direct multi-step ahead prediction model, which employs observation values and does not depend on the result of single-step prediction, provides more accurate prediction than indirect model. But in this case, the model could be asked to learn various object functions. In this paper, a hybrid model is presented based on empirical mode decomposition (EMD) and chaos analysis. The model employs EMD to decompose the original sequences into many basic modal partitions which can significantly represent potential information of original time series. And chaos features of those data sequences can be used to design DRNN. By these means, the model can be improved to learn various objective functions. And then, more precious prediction can be obtained. Finally, a benchmark time series is tested to display the advantage of this model.
点击查看大图
计量
- 文章访问数: 2793
- HTML全文浏览量: 36
- PDF下载量: 1809
- 被引次数: 0