奇异谱迭代插补的改进算法及其在缺损数据恢复中的应用  

Improved Interpolation Method Based on Singular Spectrum Analysis Iteration and Its Application in Missing Data Recovery

作　　者：王辉赞[1] 张韧[1] 刘巍[2] 王桂华[3] 金宝刚[1] 

机构地区：[1]解放军理工大学气象学院海洋与空间环境系 [2]西南交通大学信息科学与技术学院,成都610031 [3]国家海洋局第二海洋研究所,杭州310012

出　　处：《应用数学和力学》2008年第10期1227-1236,共10页Applied Mathematics and Mechanics

基　　金：国家重点基础研究计划资助项目(2007CB816003);LASG开放课题资助项目

摘　　要：基于奇异谱分析(SSA)迭代插补的基本思想,针对该方法在参数确定(主成分个数K和嵌入维数M)上存在较大随意性和计算效率低等缺陷,提出了一种参数优化的新算法——区间四分法,该方法在误差曲线存在局部波动情况下仍能有效搜索到全局最优参数解,且迭代插补精度和计算效率可得到显著提高.采用外逸长波辐射(OLR)逐日资料进行的插补试验对比分析表明,基于区间四分法改进的SSA迭代插补方案的缺损数据恢复的效率和精度良好.A novel algorithrm called interval quartering algorithm was proposed to improve the insufficiency of the conventional singular spectrum analysis iterative interpolation on parameter selection （including the number K of principal component and the embedding dimension M ）. Based on the improved singular spectrum analysis iterative interpolation, the interpolated test and comparative analysis was carried out to the outgoing longwave radiation daffy data. The results show that interval quartering algorithm can not only find the global optimal parameter to the error curve which has local oscillalion effectively but also has the advantage of fast computing speed, and this improved interpolation method is very effective to the interpolation of missing data.

关 键 词：奇异谱分析 外逸长波辐射 缺损数据插补 区间四分法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP391[自动化与计算机技术—计算机科学与技术]