Principal Component Analysis Algorithm Based on Mutual Information Credibility

Abstract

Data dimensionality reduction is the mapping of data sets from high-dimensional feature space to low-dimensional feature space. Traditional principal component analysis (PCA) algorithm is a kind of commonly used data dimensionality reduction algorithm. The computation time of the traditional PCA is usually quite long that it cannot meet the requirement of the classification for real world problems. Through the study of mutual information credibility, a PCA data dimensionality reduction algorithm based on mutual information synthesize credibility is proposed. Firstly, the ideas of mutual information (MI), relative mutual information credibility (MIR) and absolute mutual information credibility (MIA) are introduced. Then the mutual information synthesize credibility (MIS) is solved according to MIA and MIR, and feature selection is carried out by using the mutual information synthesize credibility. Finally, PCA algorithm is used to reduce the dimension of the processed data, and KNN and SVM algorithms are used to classify the data after dimensionality reduction. Compared with traditional PCA and PCA based on information entropy (E-PCA) algorithms, the results indicate that the proposed algorithm can achieve a better effect on the dimensionality reduction result and has improved classification accuracy.