To integrate global structure information into feature selection and improve the effectiveness of feature selection mechanism , an unsupervised feature selection algorithm based on low‐rank score (LRS) was presented .The low‐rank representation model with“cleanly”dictionary constraint was constructed to gain a weight matrix that with the capacities of capturing the global struc‐ture information ,identifying and expressing the data information in it .The weight matrix was introduced into the Laplacian score ,and the low‐lank score for feature selection was studied .Experimental results of data clustering and classification on public dataset verify the effectiveness of the proposed method ,which also shows that it outperforms state‐of‐art feature selection approaches .%为将数据的全局结构信息引入特征选择中，提升特征评分机制的有效性，提出一种基于低秩评分的非监督特征选择算法。利用“干净”字典约束的低秩表示模型，获得权值矩阵，该权值矩阵能够揭示数据全局结构信息，具有一定的鉴别能力，将其引入拉普拉斯评分机制，构建低秩评分机制，用于数据的特征选择。在不同的数据库上进行聚类和分类实验，实验结果表明，同传统的特征选择算法相比，该算法的性能更优。
展开▼
