闫恒乾,洪梅,张韧,朱伟军,马晨晨,余丹丹.基于偏最小二乘回归的冬季北太平洋风暴轴指数的特征诊断.气象科学,2018,(5):596-605 YAN Hengqian,HONG Mei,ZHANG Ren,ZHU Weijun,MA Chenchen,YU Dandan.Diagnosic prediction of winter storm track indices in the North Pacific based on partial least square regression.Journal of the Meteorological Sciences,2018,(5):596-605
基于偏最小二乘回归的冬季北太平洋风暴轴指数的特征诊断
Diagnosic prediction of winter storm track indices in the North Pacific based on partial least square regression
投稿时间:2016-04-08  修订日期:2016-06-14
DOI:10.3969/2016jms.0053
中文关键词:  风暴轴指数  偏最小二乘回归  核偏最小二乘回归  特征诊断
英文关键词:Storm track indices  Patial least square regression  Kernel partial least square regression  Diagnosic analysis
基金项目:公益性行业(气象)科研专项(GYHY201306028);国家自然科学基金资助项目(41575070)
作者单位E-mail
闫恒乾 国防科技大学 气象海洋学院 海洋科学与技术系, 南京 211101  
洪梅 国防科技大学 气象海洋学院 海洋科学与技术系, 南京 211101 flowerrainhm@126.com 
张韧 国防科技大学 气象海洋学院 海洋科学与技术系, 南京 211101  
朱伟军 南京信息工程大学气象灾害教育部重点实验室, 南京 210044  
马晨晨 连云港市气象局, 江苏 连云港 222006  
余丹丹 中国人民解放军61741部队, 北京 100081  
摘要点击次数: 54
全文下载次数: 27
中文摘要:
      针对现有风暴轴指数分析大多采用相关分析等较为简单方法,难以对风暴轴指数变化有效诊断分析的问题,引入偏最小二乘回归(Partial Least Square Regression,PLS)的线性方法和核偏最小二乘回归方法(Kernel Partial Least Square Regression,KPLS),对冬季北太平洋风暴轴指数变化进行了特征诊断研究,并与传统的线性无偏最小二乘回归结果进行了试验比对。结果表明:偏最小二乘回归方法的诊断结果能够更好地反映风暴轴内部变化规律,并有效降低诊断误差。对于PNYI(北太平洋风暴轴纬度指数),采用r>0.2的因子筛选方案(r为因子与风暴轴指数的相关系数)并应用KPLS算法时,预测效果最佳;对于PNXI(北太平洋风暴轴经度指数)和PNⅡ(北太平洋风暴轴强度指数),采用全因子方案并应用KPLS算法时,预测效果最佳。
英文摘要:
      With most of the existing analysises about storm track indices using such simple methods as correlation analysis and doing a bad job in diagnosic prediction, partial least square regression (PLS) containing the linear method as well as kernel partial least square regression (KPLS) is introduced to study the variation of winter storm track indices in the North Pacific. The experiment results, comparing with those of conventional least square regression, indicates a better performance of PLS in revealing the inherent variation rules and reducing the diagnosic error. In terms of the winter North Pacific storm track latitude index (PNYI), the combination of KPLS and the factor-filtering scheme of r>0.2 (r is the correlation coefficeient between the factors and the storm track indices) performs best. As for the longitude index (PNYI) and the intensity index (PNⅡ), the combination of KPLS and the all-factor scheme performs best.
查看全文  查看/发表评论  下载PDF阅读器
关闭
function PdfOpen(url){ var win="toolbar=no,location=no,directories=no,status=yes,menubar=yes,scrollbars=yes,resizable=yes"; window.open(url,"",win); } function openWin(url,w,h){ var win="toolbar=no,location=no,directories=no,status=no,menubar=no,scrollbars=yes,resizable=no,width=" + w + ",height=" + h; controlWindow=window.open(url,"",win); } &et=BE3F65FBEA4F9D3C9FA7141D352F06C249C0B0D8211F1B97F1AB27A4437213887981347E30E50FE83DE87A83EE9FD5D233FA7AB7BDECCD2BABC70124159BCA38E34C8230ECBB5823C32263521B55069BC7215C3FA14BCFD33155BED80980517A15B040E3C7EAC37CE5271FAC6D7E4040130DC482A057B122&pcid=E62459D214FD64A3C8082E4ED1ABABED5711027BBBDDD35B&cid=28A2F569B2458C17&jid=A26076E12D625EFF4F60DF07F22AE4EC&yid=EA357AD73C8E13BC&aid=&vid=&iid=94C357A881DFC066&sid=6341CCF6B158C5F9&eid=10A39635766FF5D0&fileno=20180504&flag=1&is_more=0">