Research on the rough set attribute reduction algorithm based on significance of attributes

摘要

Rough set is an effective tool to analyse and process the imprecise, inconsistent and incomplete information. Attribute reduction is the core content of rough set theory. Because of the objective of rough set data processing, it is sensitive to the noise data. Especially for large data sets processing, the decision rules obtained based on traditional rough set algorithms exhibit incompatibility and the decision information became variation. The paper proposed an improved heuristic attribute reduction algorithm and discussed the structure of heuristic information method. Using the properties of information entropy as the measure, the attribute significance is determined. By extending the concept of maximum distribution attribute reduction, the paper proposed the optimal maximum distribution attribute reduction algorithm based on attribute significance. Finally, based on the flame stability judgement of digital image, the improved algorithm is verified to be effective.