粗糙集中的离散化要求在保持原有决策系统的不可分辩关系情况下,用尽量少的断点进行离散化,而求取连续属性值的最优断点集合是一个NP难题。把连续属性值离散化问题作为一种约束优化问题,采用一种改进的遗传算法来获得最优解,并针对离散化问题设计了相应的编码方式和交叉方法。实验结果表明,采用改进的遗传算法求解连续属性值最优断点集合是可行的。
The discretization in the rough set requires that it should be maintained indiscemibility of the original decision-making system, use possible minimum number of breakpoints to discrete, and acquiring an optimal breakpoint set of continuous attributes value is a NP problem. The discretization problem of continuous attribute value as a constrained optimization problem, adopted an improved genetic algorithm to obtain the optimal solution, and designed the corresponding coding and cross method for the issue of discretization. Theex- perimental results show that the adoption of improved genetic algorithm for optimal breakpoint set of continuous attribute value is feasible.