Two-step Attribute Reduction Method Based on Fuzzy Rough Sets Dependency

To acquire the minimum attribute reduction of the dataset with continual attribute value,a two-step reduction method is proposed based on fuzzy rough sets.The concept of dependency is extended and the dependency relationship between condition attributes can be measured on the basis of the extended concept.The candidate attributes are first selected by calculating the attribute importance.Then,the candidate attributes are reduced using the dependency between attributes and of the single attribute importance.The redundant attribute is reduced via this operation.Experimental results show that the strategy can efficiently reduce the attribute dimension without sacrificing the classification accuracy.