将多尺度思想引入粗糙集连续属性离散化中,采用树结构对连续属性进行多尺度离散区间描述。发展多尺度遗传算法进行尺度优化,提出与传统遗传算法不同的交叉和变异策略。提出的方法实际上模拟了人类从不同立场、不同层次上观察和分析问题的方式,模拟人们在解决关系复杂、难于确定的问题时,由粗到细、由细到粗,逐步、反复尝试达到优化的目标。
Multiscale method is used to discretization of continuous attribute values in rough set model on trees, and a multiscale genetic algorithms is developed and its genetic operations are studied to optimize the discretization. The presented muhiscale ap- proach can represent the discretizating process in a manner from coarse to fine smoother or from fine to coarse filter as the human's intelligent behaviors.