| 季节性时间序列多步向前外推预测的研究及应用 |
| Alternative Title | Study and Application of Multi-step-ahead Extrapolation Forecast of Seasonal Time Series
|
| 赵伟刚 |
| Thesis Advisor | 李廉
; 王建州
|
| 2015-05-23
|
| Degree Grantor | 兰州大学
|
| Place of Conferral | 兰州
|
| Degree Name | 博士
|
| Keyword | 多输出预测
时变权重组合预测
高阶Markov链模型
动态预测
窗口间加权平均
|
| Abstract | 时间序列的多步向前外推预测具有极大的现实意义，本文致力于在不同预测场景下为季节性时间序列开发有效的多步向前外推预测模型：（1）对于变化趋势大体清晰但存在不规则波动成分的季节性时间序列，结合信号滤波技术和季节指数方法，提高多输出神经网络的预测准确度；（2）组合多个季节模型以充分利用它们的预测信息，同时推广一个高阶Markov链模型来更新其组合权重；（3）对于一个不断增量且在增量过程中易发生均值突变的双重季节性时间序列，建立了几个窗口间加权平均模型来开展动态预测，实现在每次数据增量后都能对未来一个小周期做出有效的预测。最后，本文将所提出的模型应用到一些预测领域中以验证它们的有效性并拓展其实际价值。实验结果表明，相对于传统预测模型，它们的预测性能都有较大提高。 |
| Other Abstract | Multi-step-ahead extrapolation forecast of time series has great practical significance.This thesis is dedicated to developing effective multi-step-ahead extrapolation forecast model for seasonal time series in various scenarios:(1) For a seasonal series with generally clear trend and slight irregular fluctuation, improve the forecast accuracy of multi-output neural network by performing signal filtering and seasonal index method. (2) Combine multiple seasonal models to fully utilize their forecast information, and generalize a high-order Markov chain model to update the combining weights. (3) For an incremental time series with double seasonality feature which possibly occurs abrupt changes in its incremental process,establish some weighted average across windows models to achieve dynamic forecast that performs an effective one-little-cycle-ahead extrapolation forecast after each data increment. Finally, this thesis applies the proposed models to some forecast fields to verify their effectiveness and expand their real value. Experimental results show their forecast performance has great improvement compared with traditional models. |
| URL | 查看原文
|
| Language | 中文
|
| Document Type | 学位论文
|
| Identifier | https://ir.lzu.edu.cn/handle/262010/225027
|
| Collection | 数学与统计学院
|
Recommended Citation
GB/T 7714 |
赵伟刚. 季节性时间序列多步向前外推预测的研究及应用[D]. 兰州. 兰州大学,2015.
|
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
