线指数特征空间内恒星光谱离群数据挖掘与分析
发布时间：2019-08-01 07:18
【摘要】：大规模光谱巡天将产生海量的光谱数据,为搜寻一些奇异甚至于未知类型的光谱提供了机会,对这些特殊天体的研究有助于揭示宇宙的演变规律和生命起源,巡天数据的离群数据挖掘有助于这些特殊的光谱的发现。利用线指数对光谱数据进行降维能够在尽可能多的保留光谱物理特征的同时,有效解决高维光谱数据聚类分析中运算复杂度较高的问题。提出了基于线指数特征的海量恒星光谱离群数据挖掘及分析的方法,以恒星光谱的Lick线指数作为光谱数据的特征,利用聚类搜寻离群数据的方法在海量光谱巡天数据搜寻离群数据,以此为基础并给出线指数特征空间内离群光谱数据的分析方法。实验结果证明:(1)以线指数作为光谱的特征值能快速的完成对高维光谱数据的离群数据挖掘,可以解决高维光谱数据运算复杂度高的问题;(2)该方法是在聚类结果上进行的离群数据挖掘,能够有效的挖掘出数量较少的发射线恒星、晚M型恒星、极贫金属星、缺失数据光谱等数据;(3)线指数特征空间的离群数据挖掘可以得到线指数特征空间内特殊恒星的发现规则。本文所提出的基于线指数特征的离群数据挖掘及分析方法可以应用到巡天数据的相关研究中。
[Abstract]:Large-scale spectral survey will produce a large number of spectral data, which provides an opportunity to search for some strange and even unknown types of spectra. The study of these special celestial bodies is helpful to reveal the evolution of the universe and the origin of life, and the mining of outlier data of patrol data is helpful to the discovery of these special spectra. Using line index to reduce the dimension of spectral data can effectively solve the problem of high computational complexity in clustering analysis of high-dimensional spectral data while retaining as many spectral physical features as possible. A method of mining and analyzing massive star spectral outlier data based on line index feature is proposed. The Lick line index of star spectrum is taken as the feature of spectral data, and the method of clustering searching outlier data is used to search for outlier data in massive spectral patrol data. On this basis, the analysis method of outlier spectral data in the feature space of line index is given. The experimental results show that: (1) using line index as the characteristic value of the spectrum can quickly complete the outlier data mining of high dimensional spectral data, which can solve the problem of high complexity of high dimensional spectral data. (2) this method is an outlier data mining on clustering results, which can effectively mine a small number of emission stars, late M stars, extremely poor metal stars, missing data spectrum and so on. (3) the discovery rules of special stars in line exponential feature space can be obtained by mining outlier data in line exponential feature space. The outlier data mining and analysis method based on line index features proposed in this paper can be applied to the related research of sky patrol data.
【作者单位】： 山东大学(威海)机电与信息工程学院;中国科学院光学天文重点实验室国家天文台;
【分类号】：P145.4
本文编号：2521607
[Abstract]:Large-scale spectral survey will produce a large number of spectral data, which provides an opportunity to search for some strange and even unknown types of spectra. The study of these special celestial bodies is helpful to reveal the evolution of the universe and the origin of life, and the mining of outlier data of patrol data is helpful to the discovery of these special spectra. Using line index to reduce the dimension of spectral data can effectively solve the problem of high computational complexity in clustering analysis of high-dimensional spectral data while retaining as many spectral physical features as possible. A method of mining and analyzing massive star spectral outlier data based on line index feature is proposed. The Lick line index of star spectrum is taken as the feature of spectral data, and the method of clustering searching outlier data is used to search for outlier data in massive spectral patrol data. On this basis, the analysis method of outlier spectral data in the feature space of line index is given. The experimental results show that: (1) using line index as the characteristic value of the spectrum can quickly complete the outlier data mining of high dimensional spectral data, which can solve the problem of high complexity of high dimensional spectral data. (2) this method is an outlier data mining on clustering results, which can effectively mine a small number of emission stars, late M stars, extremely poor metal stars, missing data spectrum and so on. (3) the discovery rules of special stars in line exponential feature space can be obtained by mining outlier data in line exponential feature space. The outlier data mining and analysis method based on line index features proposed in this paper can be applied to the related research of sky patrol data.
【作者单位】： 山东大学(威海)机电与信息工程学院;中国科学院光学天文重点实验室国家天文台;
【分类号】：P145.4
【相似文献】
相关会议论文 前2条
1 吉根林;孙志挥;;分布式系统中离群数据间相似性挖掘[A];第二十一届中国数据库学术会议论文集（技术报告篇）[C];2004年
2 金义富;朱庆生;;离群数据关键域子空间实时搜索算法[A];第二十三届中国数据库学术会议论文集（技术报告篇）[C];2006年
相关硕士学位论文 前10条
1 潘险险;基于实测运行数据的风电场仿真模型的研究[D];华北电力大学;2015年
2 蔡博文;高维数据集中离群数据挖掘方法的研究[D];合肥工业大学;2006年
3 蒋义勇;基于约束概念格的离群数据挖掘方法及应用[D];太原科技大学;2007年
4 孙伟伟;基于基尼指标和子空间的离群数据挖掘方法研究[D];太原科技大学;2012年
5 钟洵;谱聚类在离群数据挖掘中的应用研究[D];重庆大学;2010年
6 韩波;网格环境下的分布式离群数据挖掘方法研究[D];太原科技大学;2010年
7 魏翔;基于离群数据挖掘的电子商务推荐系统研究[D];武汉纺织大学;2013年
8 李永红;相关子空间中的局部离群数据挖掘及应用[D];太原科技大学;2014年
9 纪勇;离群数据挖掘在犯罪通讯痕迹发现中的应用[D];同济大学;2008年
10 刘书志;基于密度的局部离群数据挖掘方法的研究和改进[D];重庆大学;2014年
，本文编号：2521607
本文链接：https://www.wllwen.com/kejilunwen/tianwen/2521607.html
