Research on Knowledge Change Rate-Based Attribute Importance Measure Method

For the problem of attribute importance measure, basing on the decision information system, and according to the characteristics and shortcomings of the existing measure modes, we discussed the associated features between the lower and upper approximation of decision classes and the knowledge in system. Then, we constructed an attribute importance measure mode based on the knowledge change rate, and analyzed the features of the constructing measure mode from different angles. Theoretical analysis and example calculation show that the established mode is a supplement and perfect for the existing measure modes, and it not only can effectively use the existed information to reveal the associated features among each attribute, but also have a good structural characteristic and strong interpretability. It has a broad application prospect among information fusion, fuzzy decision, comprehensive evaluation and so on.