A Continuous Information Attribute Reduction Algorithm Based on Hierarchical Granulation

The attribute reduction algorithms based on neighborhood approximation usually use the distance as the approximate metric. Algorithms could result in the loss of information with the same distance threshold to construct the neighborhood families of different dimension spaces. Thereby, an attribute reduction algorithm based on hierarchical granulation is proposed. This algorithm can reduce redundant attributes in the same granularity. Experimental results with UCI data sets show that the algorithm can improve the classification power, and reduce the loss of information.