Affiliations: College of Computer, National University of Defense Technology, Changsha, China
Correspondence:
[*]
Corresponding author. Ke Wang, College of Computer, National University of Defense Technology, Changsha, China. Tel.: +86 0731 84575803; Fax: +86 0731 84575802; E-mail: [email protected].
Abstract: Numerical relationships of multi-feature data are widely concerned in data preprocessing, but semantic interpretation of the features received less attention. We completely approve of the importance of numerical relationships. However, in our opinion, the interpretative relationships of the data should be important as well. In this paper, we regard the principle component analysis (PCA) as a special case of numerical relationships. We propose an interpretative division method on the PCA and its improved algorithms from an explanatory perspective. Our method integrates the numerical data analysis with the semantic understanding of the problem. Experiments are conducted on real data sets and our method demonstrates good performance and outperforms the corresponding PCA algorithms. On the real data sets of our experiments, we also find that the interpretative features with small eigenvalues are better choices than the principle components of PCA.
Keywords: Interpretative data analysis, multi-feature data analysis, feature reduction, data preprocess, principal component analysis
