摘要
在对典型的离散化方法分析的基础上,提出一种适用于粗糙集决策表的连续属性离散化处理的通用模型结构;对遗传算法的适应度线性尺度变换作改进,将模拟退火的思想引入遗传算法,提出基于遗传模拟退火算法(GASA)的数据离散化方法,并用UCI机器学习数据库中的Iris和Glass数据集进行验证.实验结果表明,离散化方法通用模型对数据的离散化过程可以有不同的性能要求,能够保证系统中的样本对决策属性的分辨关系.GASA离散化方法具有较好的离散化效果,为连续属性离散化提供一种新的思路.
A universal model framework suitable for dealing with continuous attribute discretization of rough set decision table was presented,based on the analysis of typical discretization method.The genetic simulated annealing algorithm（GASA）-based data discretization method was proposed by modifying the transformation of linear scaling of fitness in GA and introducing the thought of SA into GA.The proposed method was applied to the Iris and Glass data sets in the UCI machine learning repository to verify its validity.The results of experiments showed that the universal model framework could set different performance requirements to the discretization process and guarantee the discernible relation of the sample to decision attributes in the system.The discretization method with genetic simulated annealing algorithm exhibited a better discretization result,providing a new idea for the discretization of continuous attribute.
出处
《兰州理工大学学报》
CAS
北大核心
2011年第1期91-94,共4页
Journal of Lanzhou University of Technology
关键词
粗糙集
离散化
遗传模拟退火算法
rough sets
discretization
genetic simulated annealing algorithm
