电子工业出版社-网上书店

随着LAMOST正式巡天的实施，已成功获取600万条天体光谱以及星表，并每天以海量的数字增长着，对长期传统的人工分析、人眼证认等任务带来了巨大挑战。本书以河外星系和恒星光谱为研究背景，针对天文学研究中稀有天体的特征分析以及天体光谱的分类等任务，将新兴的数据挖掘技术应用到天体光谱规律的发现和研究中，并从天文物理学角度对挖掘结果进一步分析。主要包括稀有、离群天体光谱的搜寻与分析、天体光谱分类方法与分析两个方面的内容。

第1 章 绪论······················································································ 1
1.1 天体光谱·············································································· 1
1.1.1 LAMOST 光谱巡天·························································· 2
1.1.2 SDSS 光谱巡天································································ 5
1.1.3 光谱分析········································································ 6
1.2 数据挖掘·············································································· 7
1.2.1 产生和定义····································································· 7
1.2.2 数据挖掘任务与分类······················································ 10
1.2.3 主要应用······································································ 12
1.3 海量天体光谱数据挖掘······················································ 14
1.3.1 分类············································································ 14
1.3.2 聚类及离群分析···························································· 17
1.3.3 关联规则······································································ 19
1.3.4 恒星大气参数测量························································· 20
1第2 章 基于模糊识别的双红移系统星系光谱搜寻与分析·············· 24
2.1 引言···················································································· 25
2.2 基于模糊识别的搜寻方法·················································· 27
2.2.1 样本选择······································································ 27
2.2.2 方法描述······································································ 28
2.3 结果分析············································································ 35
2.3.1 SDSS DR9 和LAMOST DR1 中SGPs 样本························· 35
2.3.2 光谱与图像分析···························································· 39
2.3.3 尘埃消光测量································································ 48
2.4 讨论···················································································· 51
第3 章 稀有光谱检索的PU 学习方法············································ 53
3.1 问题提出············································································ 54
3.2 二部排序模型····································································· 56
3.2.1 TopPush 方法································································ 57
3.2.2 面向稀有光谱检索的BaggingTopPush 方法························ 58
3.3 实验设计············································································ 59
3.3.1 样本选择······································································ 60
3.3.2 实验设置······································································ 61
3.3.3 评价指标······································································ 64
3.4 结果分析············································································ 65
3.4.1 排序效果······································································ 65
3.4.2 排序效率······································································ 72
3.4.3 参数敏感性··································································· 74
3.5 讨论···················································································· 76
第4 章 E+A 星系搜寻与分析·························································· 78
4.1 问题提出············································································ 78
4.2 E+A 星系光谱搜寻方法······················································ 80
4.2.1 样本选择—LAMOST 数据集········································· 80
4.2.2 搜寻方法······································································ 80
4.2.3 近邻E+A 星系星表························································ 83
4.3 结果分析············································································ 87
4.3.1 样本分布特征································································ 87
4.3.2 星族合成分析································································ 90
4.3.3 图像分析······································································ 92
4.4 讨论···················································································· 95
第5 章 基于贝叶斯支持向量机的光谱分类方法····························· 98
5.1 问题提出············································································ 98
5.2 基于贝叶斯支持向量机的分类方法·································· 100
5.2.1 支持向量机································································· 100
5.2.2 贝叶斯推理································································· 101
5.2.3 马尔可夫链蒙特卡罗···················································· 101
5.2.4 贝叶斯支持向量机······················································· 102
5.3 实验分析·········································································· 107
5.3.1 样本选择···································································· 107
5.3.2 预处理方法································································· 108
5.3.3 实验参数设置······························································ 112
5.3.4 结果分析···································································· 113
5.4 讨论·················································································· 116
第6 章 基于分类模式树的恒星光谱自动分类方法······················· 117
6.1 问题提出·········································································· 118
6.2 恒星光谱分类模式树························································ 119
6.3 分类模式树构造方法························································ 120
6.3.1 算法思想···································································· 120
6.3.2 算法描述···································································· 121
6.3.3 算法分析···································································· 122
6.4 分类规则提取及恒星光谱分类········································· 122
6.5 实验分析·········································································· 123
6.6 讨论·················································································· 127
第7 章 恒星光谱分类规则后处理方法········································· 129
7.1 问题提出·········································································· 129
7.2 基于谓词逻辑的分类规则后处理方法······························ 131
7.2.1 恒星光谱分类规则······················································· 131
7.2.2 恒星光谱分类规则后处理·············································· 132
7.2.3 实验分析···································································· 136
7.3 基于集合运算的分类规则后处理方法······························ 138
7.3.1 分类规则问题描述······················································· 138
7.3.2 分类规则后处理算法···················································· 140
7.3.3 实验分析···································································· 142
7.4 讨论·················································································· 143
参考文献·························································································· 144
附录A SDSS DR9 和LAMOST DR1 的SGPs 样本清单············ 149
附录B LAMOST DR2 的E+A 样本清单······································· 159
附录C LAMOST DR2 的E+A 样本测光信息清单························ 163

前 言
仰望璀璨的星空，辽阔而深邃，自由而宁静，吸引着人们苦苦追
寻与不断探索的向往。LAMOST 是一架横卧南北方向的中星仪式反射
施密特望远镜，在5 度视场、直径为1.75 米的焦面上放置4000 根光
纤，可以同时获得4000 个天体的光谱，是当前世界上光谱获取率最高
的望远镜。随着LAMOST 正式巡天的实施，已成功获取600 万条天
体光谱以及星表，并每天以海量的数字增长着，对长期传统的人工分
析、人眼证认等任务带来了巨大挑战。而数据挖掘，作为一门新兴的
学科分支，涉及人工智能、机器学习、模式识别等多个学科领域，主
要任务是从大量的原始数据中提取潜在的、人们感兴趣的知识，已被
广泛地应用于科学、工程、商业等领域。将数据挖掘技术应用于海量
的天体光谱数据中，获取潜在的、有意义的天体规律及性质，对更有
效地使用巡天数据、进一步深入天文学理论研究都具有比较重要的应
用价值。
近年来作者一直从事数据挖掘应用与天体光谱分析交叉领域的研
究，在深入了解光谱分析任务、分析当前数据急剧增长特点的基础上，
结合计算机技术优势，开展了一系列的研究工作，本书是近年来相关
科研成果的总结。全书除绪论主要介绍天体光谱数据的主要特征以及
数据挖掘技术的基本理论外，主要内容分两大部分共6 章，具体章节
编排如下：
（一）特殊、稀有天体的挖掘及分析（包括第2 章至第4 章）。第
2 章针对星系光谱中呈现的双红移系统，提出一种基于模糊识别的光
谱特征线识别方法，并采用SDSS DR9 和LAMOST DR2 的星系光谱
数据，系统地搜寻了具有双红移系统的星系光谱并对其结果进行了光
谱及图像分类、特例分析、前景星系消光测量等方面的讨论；第3 章
针对碳星光谱中存在的模板较少从而导致从海量数据中搜寻比较困难
的问题，提出一种新的高效的PU 学习方法，并选择SDSS DR10 中十
万余条光谱实现验证了该方法的搜寻质量和效率。第4 章针对
LAMOST 河外星系光谱分辨率及信噪比等特征，修正了[OII]、H?、
H?特征线边界，通过测量其等值宽度并按照经典（Goto.提出）的判
定依据，从LAMOST DR2 中系统搜寻了E+A 星系，并对其结果进行
了红移分布、空间分布、星等分布特征、图像特征以及星族特征等方
面的讨论。
（二）光谱分类及后处理方法研究（包括第5 章至第7 章）。第5
章针对巡天数据分析中最基本的光谱型分类问题，提出一种基于贝叶
斯支持微量机的光谱自动分类方法，选择SDSS DR10 的M 型恒星光
谱，实验验证了该方法在光谱子型的分类上具有较高的准确率及效率，
同时对预处理过程中噪声、归一化方法、特征提取方法对分类结果的
影响进行了讨论；第6 章针对恒星光谱分类任务，提出一种基于分类
模式树的恒星光谱分类规则挖掘方法。采用SDSS 恒星光谱作为实验
数据，验证了该方法的正确性，而且具有较高的分类正确率；第7 章
针对采用数据挖掘方法提取的光谱分类规则中存在的冗余严重影响分
类效率和质量的问题，提出基于谓词逻辑、集合运算的两种分类规则
后处理方法，从而减小分类器的大小。采用SDSS 恒星光谱数据，实
验验证了这两种方法在不降低分类准确率的前提下，可以有效提高分
类效率。）特殊、稀有天体的挖掘及分析（包括第2 章至第4 章）。
本书的完成得到了太原科技大学人工智能实验室、计算机科学与
技术学院、中科院国家天文台各位老师的大力支持，特别是张继福教
授、罗阿理研究员为本书提出了许多宝贵的建议，在此一并致以诚挚
的谢意。
本书所涉及的部分研究工作得到了国家自然科学基金项目（项目
编号： 61272263， 61572343）、山西省科技攻关项目（ 项目编号：
2015031009）和太原科技大学博士启动基金（项目编号：20162007）
的资助，在此向相关机构表示深深的感谢。
由于作者水平有限，书中难免有不妥之处，欢迎各位专家和广大
读者批评指正。
编 者
2016 年11 月

科技/计算机/数据库

杨海峰，博士，太原科技大学计算机学院副教授，研究方向：人工智能与数据挖掘、河外星系光谱分析。近年来主持和参与国家级、省部级科研项目多项，发表学术论文近30多篇，其中SCI收录18篇。

无

本书配套资源下载

对不起，暂无音视频资源！

天体光谱数据挖掘与分析