摘要
以分类为基础提出了一种基于条件熵的决策表属性并行约简算法。该算法通过条件熵的计算在属性约简的同时将原决策表逐层分解为相对于决策属性来说尽量均匀的子决策表,从而实现了属性约简的并行计算。随后对该算法的时间复杂度进行了分析,实验表明,该算法在效率方面优于传统算法。
This paper proposed a parallel algorithm of attributes reduction of decision table based on conditional entropy.Through the calculation of the entropy,this algorithm which was on basis of classification reduced the attributes while it was disassembling the decision table into sub-tables as symmetrical as possible correspondingly,so the algorithm achieved a parallel calculation of attribute reduction.Also analyzed the time complexity of the algorithm.The experiment shows that the new algorithm is more effective than traditional algorithm.
出处
《计算机应用研究》
CSCD
北大核心
2011年第4期1296-1298,1321,共4页
Application Research of Computers
基金
甘肃省科技支撑计划资助项目(1011GKCA040)
兰州市企业技术攻关计划资助项目(2009-1-4)
兰州交通大学"青蓝"人才工程基金资助项目(QL-05-10A)
关键词
决策表
属性约简
条件熵
并行计算
decision table
attributes reduction
conditional entropy
parallel algorithm
