An improved PCA algorithm based on WIF

Fengxiang Jin, Shifei Ding

Abstract: In this paper, we analyze the information feature of principal component analysis (PCA) deeply based on information entropy. According to idea of entropy function, a new weighted information functions (WIF) is proposed, and the information content of data matrix X is measured by it. Based on WIF, the information compression rate (ICR, RIC) and accumulated information compression rate (AICR, RIC) are set up, by which the degree of information compression is measured. At last, an improved PCA algorithm (IPCA) based on WIF is constructed. Through simulated application in practice, the results show that the IPCA proposed here is efficient and satisfactory. It provides a new research approach of feature compression for pattern recognition, machine learning, data mining and so on.