提出和探讨了一种新的基于模糊粗糙集和断点简约化的离散化方法.综合考虑到规则的支持度和可信度及其关系,应用属性离散指标作为离散化的标准,证明了该指标可以作为离散化彻底的充分条件.并且在时间复杂度和空间复杂度方面分析了算法的有效性,与同类算法比较可以发现该算法在基本不损失分类信息的基础上有效降低这两方面的复杂度,能有效地避免以往各种算法中出现的弊端.最后将其应用于电网故障诊断中,通过具体算例测试,证明该算法的有效性和实用性.
A new discretization algorithm based on fuzzy rough set with the number of breakpoints reduced is put forward and discussed, taking account of the supportability and confidence of its rules and the relation between them. Taking the discretization indices of attributes as criteria, the indices are proved the sufficient conditions for complete discretization. Discussers the effectiveness of the algorithm in view of the time and space complexity and compares the algorithm with similar ones. It is found that the algorithm can reduce the time and space complexity efficiently, thus avoiding the drawbacks often found in other conventional algorithms. Its effectiveness and practicality are verified through a numerical example applying it to fault diagnose for power network.