一种错误率可控的混沌时间序列区间预测算法
CSTR:
作者:
作者单位:

(天津大学电气自动化与信息工程学院,天津300072)

作者简介:

王迪(1989-), 男, 博士生, 从事模式识别、神经网络、一致性预测的研究;王萍(1955-), 女, 教授, 博士生导师, 从事模式识别、 图像识别、 运动对象跟踪等研究.

通讯作者:

E-mail: wangdi2015@tju.edu.cn.

中图分类号:

TP273

基金项目:

天津市青年基金项目(2016FH-0011).


An interval prediction algorithm for chaos time series with controllable error rate
Author:
Affiliation:

(School of Electrical and Information Engineering,Tianjin University,Tianjin300072,China)

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对高风险背景下的混沌时间序列区间预测问题,首次将回声状态网络与一致性预测框架相结合,提出基于两者的混沌时间序列区间预测算法.该算法将回声状态网络的拟合能力与一致性预测区间的可靠性相结合,使得最终的预测区间包含被预测值的频率或概率可以被显著性水平参数所控制,即预测区间具有极高的可信度.同时,由于使用岭回归学习回声状态网络的输出权重,使得算法在学习阶段对样本的留一交叉估计可以被快速地计算,极大地缩短了一致性预测的学习时间.理论分析表明,所提出算法的时间复杂度等价于原始回声状态网络算法的时间复杂度,即算法具有较快的计算速度.实验表明,所提出算法能够较精确地控制预测的错误率,对噪声具有鲁棒性,且预测区间比基于高斯过程的预测区间更加准确地刻画了被预测值的波动范围.

    Abstract:

    To tackle the chaos time series interval prediction problem with high risk, this paper proposes an algorithm combining the ideas of echo state network and conformal prediction. While inheriting the learning ability from echo state network, the algorithm is capable of predicting reliable intervals for chaos time series with the help of the learning framework of conformal prediction, as the error rate of the prediction intervals can be controlled by the preset parameter named significance level. Meanuhice, due to the use of ridge regression to train the output weights of echo state network, the leave-one-out estimates in the training set can be computed efficiently, which accelerates the learning process of the algorithm effectively. As such, from the point of view of algorithm complexity, it is shown that the time complexity of the whole algorithm is equivalent to that of echo state network, implying that the algorithm is as fast and of practical use as echo state network. The experimental results show that the proposed algorithm is empirically valid as an interval predictor, robust to noise, and the prediction intervals output by the proposed algorithm are more related to the variance of the predicted values than those output by Gaussian process.

    参考文献
    相似文献
    引证文献
引用本文

王迪,王萍,石君志.一种错误率可控的混沌时间序列区间预测算法[J].控制与决策,2019,34(5):956-964

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2019-04-17
  • 出版日期:
文章二维码