摘要
鉴于现有属性约简算法大多是针对一致决策表不适合对不一致决策表的求解,且运行效率底.文章提出了一种新的决策表属性约简算法.首先判断决策表是否为一致决策表;如果是不一致决策表,则通过最大相对概率分布将其转换为一致决策表;然后基于一致决策表的相对概率分布满足单点分布的性质来快速确定属性是否为冗余属性,从而求出约简;并通过MATLAB和UCI学习数据库验证了该算法的有效性和高效性.
The existing attribute reduction algorithm is mostly based on consistent decision tables. It is not suitable for solving inconsistent ones and also its operating efficiency is inefficiency. A new algorithm is putted forward. Firstly the algorithm judges whether the decision table is consistent, if it is inconsistent, using the maximum relative probability distribution to turn the inconsistent decision table into consistent decision table. Then bases on the relative probability distribution of consistent decision table are satisfied one-point distribute to judge whether the attribute is redundant attribute, and thus obtains reduction. The validity and feasibility of this algorithm are demonstrated by MATLAB and UCI machine learning databases.
出处
《太原师范学院学报（自然科学版）》
2009年第3期46-49,56,共5页
Journal of Taiyuan Normal University:Natural Science Edition
基金
漳州师范学院科研基金(SK08004)
关键词
粗糙集
属性约简
相对概率分布
最大分布约简
单点分布
rough set
attribute reduction
relative probability distribution
maximal distributive reduction
one-point distribute
