| 时序数据的非线性最小二乘迭代分解算法 |
投稿时间：2018-03-01 修订日期：2018-04-09  点此下载全文 |
| 引用本文： |
| 摘要点击次数: 103 |
| 全文下载次数: 0 |
|
| 基金项目:广东省应用型科技研发专项基金资助项目（2015B010130003）、佛山职业技术学院校级重点科研项目（No.2015KY006）、佛山职业技术学院2018年横向重点资助项目《时序数据的高效辨识算法及应用》 |
|
| 中文摘要:从时序数据中精确地分解出趋势、周期及随机噪声等数据成分，能有助于人们掌握事物在演变过程中所蕴藏的内在规律，基于非线性最小二乘法，提出一种性能更为高效的时序数据分解算法。首先，基于关键转折点和趋势导数的方法从待分解序列中概要地析出各种不同的数据成分，然后，分别利用多项式函数、正弦谐波级数及自回归模型对相应的数据成分进行拟合，最后，在加法模型中迭代求解各种数据成分的非线性最小二乘参数。实验表明，新设计的算法在分解精度和计算成本等指标上均优于现有的算法。 |
| 中文关键词:时序分解 非线性最小二乘 关键转折点 趋势导数 |
| |
| Nonlinear least squares iterative decomposition algorithm for time series data |
|
|
| Abstract:Precisely the decomposition trend, cycle and random noise data components from time series data, can help people grasp the things inherent in the laws contained in the evolution process, based on the nonlinear least squares method, proposes a performance more efficient sequential data decomposition algorithm. First of all, the key turning point method and trend based on derivative from the precipitation of various components in the sequence data, and then briefly, respectively using the polynomial function, sinusoidal harmonic series and autoregressive model of the corresponding data elements were fitted. Finally, iterative solution of various data component parameters in the nonlinear least squares model in addition. Experimental results show that the new algorithm outperforms the existing algorithms in the decomposition accuracy and computation cost and other indicators. |
| keywords:time series decomposition non-linear least squares key turning point trend derivative |
| 查看全文 查看/发表评论 下载pdf阅读器 |
位访问者,今日访问
次