A new method for feature extraction - Class-wise non-locality preserving projection

作者：Wang Wen Jun^*; Zhang Jun Ying

来源：Transactions of Beijing Institute of Technology, 2010, 30(SUPPL. 1): 128-131.

• CLPP
• Clustering feature
• Feature extraction methods
• Gene Expression Data
• Locality preserving projections
• Nonlocalities

In contrast to the existing class-wise locality preserving projection (CLPP), a technique based on the characterization of the inner-cluster scatter, a novel feature extraction method for gene expression data, called class-wise non-locality preserving projection (CNLPP), is proposed, which is based on the characterization of the inter-cluster scatter. The idea of CNLPP was realized with Matlab to show the effectiveness of visualization and clustering recognition on gene expression data after feature extraction and was compared with PCA, CLPP. The results indicate that CNLPP is more effective for feature extraction than CLPP when the inter-cluster information plays a dominant role in discrimination. The new method is more suitable for use in the task of classification when the inter-cluster information plays a dominant role.

• 出版日期2010
• 单位西安电子科技大学