1. College of Mathematics and Computer Science, Guangdong Ocean University, Zhanjiang 524088;2. School of Mathematical Sciences, Beijing Normal University, Beijing 100875;3. School of Materials Science and Engineering, Hunan University of Science and Technology, Xiangtan 411201;
巡天观测与高能物理、黑洞天文等领域均有密切的联系. 基于星系-超新星二分类问题, 研究光谱数据预处理, 结合余弦相似度改善PCA (Principal Component Analysis)光谱分解特征提取方法, 用SDSS (the Sloan Digital Sky Survey)、WISeREP (the Weizmann Interactive Supernova data REPository)组成的5620条光谱数据集训练支持向量机, 可以得到0.498%泛化误差的识别模型和新样本分类概率. 使用Neyman-Pearson决策方法建立NPSVM (Neyman-Pearson Support Vector Machine)模型可进一步降低超新星的漏判率.
Sky survey is closely related to the developments of many domains such as high energy physics and black hole astrophysics. In order to solve the classification problem between galaxy and supernova, an available supernova recognition method based on NPSVM (Neyman-Pearson Support Vector Machine) has been proposed. The dataset, which is collected from WISeREP (the Weizmann Interactive Supernova data REPository), SDSS (the Sloan Digital Sky Survey) and supernova templates made by Nugent, has 3427 supernova spectra and 2193 galaxy spectra. After preprocessing spectral data, the decomposed spectrum feature based on the Principal Component Analysis (PCA) is extracted, and the redundant features are decreased with the cosine similarity method. The classification model based on Support Vector Machine (SVM) has a low level of generalization error evaluated 0.498%, and can calculate the classification probability for a new sample. Furthermore, the improved NPSVM model can limit the missing rate on supernovae with the Neyman-Pearson criterion.