利用基于互信息的不平衡分类方法识别稀有光谱（英文）

发布时间：2018-07-07 10:08

本文选题：不平衡分类 + 互信息　；参考：《光谱学与光谱分析》2016年11期

【摘要】：从海量恒星光谱中发现稀有光谱是天文学研究的重要课题之一。与一般光谱相比,稀有光谱数量较少,因此,传统分类方法无法正常工作。究其原因是这些方法不仅在分类决策时并未对稀有光谱予以更多关注,而且只关注分类的准确率。鉴于此,在总结当前分类方法的基础上,深入分析互信息与决策树之间的关系,提出基于互信息的代价缺失决策树。SDSS DR8中K型、F型、G型以及M型恒星光谱上的比较实验表明,与传统分类方法相比,所提方法能够较好地完成稀有光谱识别的任务。
[Abstract]:The discovery of rare spectra from massive stellar spectra is one of the most important topics in astronomical research. Compared with the general spectrum, the number of rare spectra is less, so the traditional classification method can not work properly. The reason is that these methods not only do not pay more attention to rare spectrum but also to the accuracy of classification. In view of this, on the basis of summing up the current classification methods, the relationship between mutual information and decision tree is analyzed in depth, and the comparison experiments of K F type G type and M type star spectrum in SDSS DR8 based on mutual information cost missing decision tree are presented. Compared with the traditional classification method, the proposed method can accomplish the task of rare spectrum recognition.
【作者单位】：中北大学软件学院;中国科学院国家天文台光学天文重点实验室;中国科学院国家天文台;
【基金】：Scientific and Technological Innovation Programs of Higher Education Institutions in Shanxi(2014142) Program for the outstanding Innovative Teams of Higher Learning Institutions of Shanxi
【分类号】：P144.1

【相似文献】