PRINCIPAL COMPONENT ANALYSIS ALGORITHM BASED ON FUZZY SIMILARITY MEASURE
It is well-known that principal component analysis algorithms are proposed in order to reduce the dimension of higher-dimensional data, by which, the feature dimension of a dataset can be reduced. Through abstracting the principal components, the new variables, which are less dimensional and mutually independent, could be used to represent the most information provided by the original ones, and then do a further analysis (e.g.,principal component regression, principal component clustering and feature extraction.) on the low-dimensional data consisting of the principal components. However, the basic thought of principal component analysis is to build a few linear combinations of the original variables, where each principal component merely represents the linear dependence among feature vectors in the original dataset. Obviously, if there exist non-linear relations among the feature vectors in the given dataset, the validity of the principal component analysis will be reduced drastically. Consequently, in order to reduce the dimension of the dataset, in which non-linear dependence exists among the features, a principal component analysis algorithm based on fuzzy similarity measure is proposed in this paper. In the algorithm, using the matrix of fuzzy similarity measure as a substitute of covariance matrix;the principal components of the dataset, in which non-linear dependence exists, are extracted, and hence the information provided by the new variables is more sufficient than the one in the classical principal component analysis algorithms.
Fuzzy similarity measure Principal component analysis Dimension reduction
NAXIN CHEN ZHUOMENG ZHANG
Department of Applied Mathematics in Dalian Maritime University, Dalian, Liaoning Province, China Jinzhou Hygienic School, Jinzhou 121001, P.R.China
国际会议
The Second International Conference on Information & Systems Sciences(ICISS2008)(第二届信息与系统科学国际会议)
大连
英文
954-962
2008-12-18（万方平台首次上网日期，不代表论文的发表时间）