A new attribute reduction method based on decision-theoretic rough set

Decision-theoretic rough set is a rough set model with good tolerance to noise data. However, due to the non monotonicity of the positive region, the traditional attribute reduction can not be directly constructed. In this paper, we propose a new attribute reduction method in decision-theoretic rough set. Firstly, a new definition of attribute reduction is given, that is, the positive region of attribute reduction must not be less than the positive region of the complete set of attributes, and then a corresponding reduction algorithm is proposed based on this definition. Finally, a series of simulation experiments are carried out to prove the effectiveness and superiority of the algorithm by three methods, the size of attribute reduction, the classification accuracy of attribute reduction set and the efficiency of algorithm.