Apply the attributes correlation based on information entropy to attribute reduction in random information systems

Taking random information systems as study subjects, a new attribute reduction method is proposed by using correlation between two subsets. Information entropy about logarithmic form is used to measure uncertainty of knowledge, and the correlation connection between information entropy, conditional information entropy, joint information entropy and mutual information entropy are analyzed in a random information system. Correlation coefficients are introduced to describe the correlation between two attribute subsets. Attribute correlations are introduced to depict how important the attributes. A new algorithm based on attribute correlation is given in random information systems, and the experimental result shows that the algorithms is effective.