基于二进制粒子群优化的决策系统属性离散化

Attribute Discretization for Decision System Based on Binary Particle Swarm Optimization

为解决连续属性无法直接用于粗糙集理论的问题,依据粗糙集连续属性离散化的根本要求,提出了一种基于二进制粒子群优化算法(Binary Particle Swarm Optimization,BinaryPSO)的属性离散化方法。该方法将二进制粒子视为断点子集,最小化断点集中的断点个数作为优化目标,粗糙集属性分类精度作为约束条件。其中,适应函数的定义保证了在尽量减少决策系统信息损失的前提下,得到简化的决策系统。仿真结果表明,该方法得到的离散结果包含较少的断点个数,并且保持了较高的分类能力。

To solve the problem that continuous attributes can not be used directly to rough set theory,an attribute discretization approach based on binary particle swarm optimization is presented according to the basic requirements of rough set continuous theory discretization.Considering binary particle as cut-point set,a minimal set of cut-points is determined as optimal target when rough set attribute classifying precision is taken as constraint condition.The fitness function is defined to get predigested decision system with least loss of decision system information.The simulation result shows that the obtained discrete result contains less cut-points number,and the higher classification ability is maintained.