Attribute Reduction Method Based on Binary Sparrow Search Algorithm and Neighborhood Rough Set

Abstract

In this paper, an attribute reduction algorithm of neighborhood rough set based on variable precision is proposed, and the lower approximate calculation formula is redefined, which solves the shortcomings of traditional neighborhood rough set, such as fault tolerance and weak noise resistance. Considering that the traditional attribute reduction algorithm is serial search, it is easy to fall into the local optimal solution. Therefore, this paper proposes to combine intelligent optimization algorithm with neighborhood rough set. On the basis of the excellent performance of continuous sparrow search algorithm (SSA), a binary sparrow algorithm (BSSA) is proposed to solve the binary discrete problem, which is used to solve the optimal combination of attributes. In this paper, according to the definition of positive region monotonicity, distance matrix is introduced. For the positive region sample in a single attribute, it can be omitted in the subsequent calculation, and the calculation of positive region is transformed into the form of adding several matrices, which provides an effective method to solve positive region for variable precision neighborhood rough set. Finally, five UCI standard data sets were used as experimental objects to compare with other 4 attribute reduction algorithms. The experimental results show that both the number of attribute parsimony and the accuracy of classification, the proposed algorithm shows a strong advantage.