基于相对优势类差量的序信息系统属性约简算法

Algorithm for Attribute Reduction in Ordered Information Systems Based on Relative Differential Quantity of Advantage Classes

针对序信息系统,定义了优势类差量和相对优势类差量,讨论了一种刻画属性重要度的新指标。在此基础上,提出一种基于相对优势类差量的启发式约简算法,它能从搜索空间中删除不重要属性,降低了约简的时间复杂度。与现有启发式约简算法相比,该算法具有较高的搜索效率。实例与实验表明了该算法的可行性与有效性。

The notions of a differential quantity between advantage classes and a differential quantity between relatively advantage classes are defined in ordered information systems, and a new index for measuring significance of the attributes of an ordered information system is also presented. Then, a heuristic reduction algorithm for ordered information systems is proposed. The proposed algorithm is of less time complexity since it adopts the strategy of removing unimportant attributes from the search space. Compared with the existing reduction algorithms developed for the ordered information systems, presented algorithm has certain advantages in decreasing the number of searching attributes. Finally, a real example and some experimental results are used to demonstrate the feasibility and effectiveness of the proposed algorithm.