一种晚型天体光谱离群数据挖掘系统

发布时间：2018-10-16 12:54

【摘要】：探索海量的M型恒星中具有磁活动、巨星等较特殊、稀有的天体,对于后续观测、银河系结构、演化等科学研究具有重要的意义,针对M型恒星光谱特征线出现在子空间中的局部偏离,设计并实现了晚型恒星离群数据挖掘系统。首先采用稀疏因子和稀疏度系数度量样本在属性空间上的分布特征,并在此基础上对M型恒星光谱特征线进行离散化、降维等预处理,获得光谱子空间;然后采用微粒群算法搜索离群子空间,并证认子空间内光谱是否离群;此外,选择SDSS M型光谱特征线指数集为样本,实验分析了稀疏因子和稀疏度系数的设置对离群结果的影响,并将离群挖掘结果与SDSS提供光谱型等参数对照,表明利用该系统实现晚型恒星光谱特征线局部离群数据挖掘是可行并有价值的。
[Abstract]:It is of great significance to explore the special and rare celestial bodies such as magnetic activity, giant stars and other special and rare objects in the massive M-shaped stars, which are of great significance for the scientific research of the subsequent observation, the structure of the Milky way, the evolution of the Milky way, etc. Aiming at the local deviation of the spectral characteristic line of M-type stars in subspace, a late-type star outlier data mining system is designed and implemented. Firstly, the sparse factor and the sparsity coefficient are used to measure the distribution characteristics of the samples in the attribute space, and on this basis, the spectral characteristic lines of M-type stars are discretized and reduced to obtain spectral subspace. Then the particle swarm optimization algorithm is used to search the outlier subspace and to identify whether the spectrum is outliers in the subspace. In addition, the SDSS M-type spectral feature line index set is selected as the sample. The effects of sparse factor and sparsity coefficient on outliers are analyzed experimentally, and the results of outlier mining are compared with the spectral types provided by SDSS. The results show that it is feasible and valuable to mine local outlier data of the spectral characteristic line of late star by using this system.
【作者单位】：太原科技大学计算机科学与技术学院;
【基金】：国家自然科学基金项目(41372349,61272263) 山西省自然科学基金项目(2012011011-4)资助
【分类号】：P144.1;TP311.13

【参考文献】