Fast attribute reduction algorithm based on discernibility matrices
Ping Lv, Jin Gui Qian, Bo Wang, Feng Kun Xie · Computer Engineering and Applications Journal · 2010
In order to solve the efficiency problem of calculating the attribute reduction based on discernibility matrices,the shortcomings of attribution reduction algorithm are analyzed based on discernibility matrices,and the definition of the new discernibility matrices is presented.It decreases greatly the number of non-empty elements,which improves the efficiency of algorithm for attribute reduction based on discernibility matrices.And an attribute with indiscernibility can be used to compute the attribute with more frequencies for generating smaller discernibility matrices.A new algorithm based on the improved discernibility matrices is proposed.An example is used to illustrate the efficiency of the new algorithm.The simulation experiments for UCI databases show that the new algorithm is efficient for various kinds of data sets.