基于关系矩阵的信息系统属性约简算法

Attribute Reduction Method in Information System Based on Relation Matrix

提出了信息系统属性约简的一种新算法,从关系矩阵角度定义了属性重要度及非核属性次要度,然后以属性重要度指标作为启发式信息,辅以属性次要度,设计了一种简洁、有效的属性约简算法。该算法具有较大的灵活性,能从搜索空间逐次删除重要度为零及次要度低的属性,避免对这些属性进行重复搜索,提高了搜索的效率,最后通过实例验证了算法的可行性与有效性。

A novel algorithm for attribute reduction in decision system is proposed.Firstly,attribute significance and the significance of non-nuclear attribute are defined from the perspective of relation matrix.Then the new attribute significance is used as heuristic information to design a new attribute reduction algorithm,assisted by the significance of non-nuclear attribute.Compared with the existed algorithms,the algorithm developed in this paper can avoid repeatedly searching these attributes by gradually deleting unimportant attributes from searching space.Thus the efficiency can be improved.Finally,an example was given to verify both the feasibility and efficiency of the algorithm.