基于概念格的天体光谱离群数据识别方法

A Concept Lattice Based Recognition Method of Celestial Spectra Outliers

在宇宙中,寻求特殊的、未知的天体是人类探索宇宙奥妙所追求的目标之一,天体光谱离群数据识别方法是实现该目标的有效手段之一.将概念格中每个概念节点内涵描述为天体光谱数据特征子空间,提出了一种天体光谱离群数据识别方法.首先将概念节点的内涵缩减看作天体光谱特征子空间,并依据稀疏度系数阈值确定稀疏子空间;其次对于稀疏子空间,依据稠密度系数判定祖先概念节点内涵是否为稠密子空间,进而判断出概念节点外延中包含的数据对象是否为天体光谱离群数据;最后以离散化天体光谱数据作为形式背景,实验验证了利用该方法识别出的天体光谱离群数据是准确的、完备的和有效的。

It is one of the main goals in mankind＇s universe exploration to find unknown and particular celestial bodies. Outliers recognition is an effective way of finding the spectrum data of unknown and particular celestial bodies in mass celestial body spectrum data. A recognition method of celestial spectra outliers based on concept lattice is proposed by regarding the intension of the concept lattice nodes as characteristic subspace of the celestial spectra. First, the intension reduction of the concept lattice nodes is regarded as the characteristic subspace of celestial spectra, and the sparsity subspace is defined according to the sparsity coefficient threshold. Second, whether the intension of the ancestor nodes for sparsity subspaces is a dense subspace, is decided by the dense coefficient threshold, and accordingly the extent of the node is defined as the celestial spectra outliers if the subspaces are dense. Finally, the experiment results validate the method by taking the celestial spectra as the formal context.