Abstract: In order to solve the problem of calculating the attribute reduction of high-dimensional data set, a new pseudo particle swarm based on updating the motion equation of particle swarm is proposed, and dependency of the attributes, feature number of the subset of attributions are added into fitness function. Experimental results show that the proposed algorithm can get better results than some other existing algorithms for attribute reduction in terms of both solution quality and computational effort.

Key words: rough set, attribute reduction, core attribute, particle swarm

摘要： 针对高维数据集的属性约简问题，通过改变经典粒子群算法的运动方程，并用属性依赖性和属性子集特征数构造适应度函数，提出以决策表核属性为基础的最小属性子集搜寻策略。实验结果表明，与其他类型的最小属性约简算法相比，该算法不仅能有效提高获得最小属性约简的机率，同时还大大降低了计算时间。

关键词: 粗糙集, 属性约简, 核属性, 粒子群