摘要
针对传统基于差别矩阵的不一致决策表规则获取算法效率不理想的情况,提出了一种快速的基于差别矩阵的规则获取算法。算法首先引入简化决策表思想,删除决策表中可能存在的许多重复对象;然后基于简化决策表构造不同决策类之间的子差别矩阵,以有效地解决对象分布的非平衡性问题和缩小算法的求解空间;且采用启发式向后贪心搜索策略求解相对最小属性约简;并根据规则可信度获取有效的决策规则,可信度可动态设置,使算法具有较好的适应性。最后通过算例分析和实验比较验证了算法能获取有效的决策规则。
Since the efficiency of traditional rule extraction algorithms based on discernibility matrix in inconsistent deci- sion table is often poor, a quick rule extraction algorithm based on discernibility matrix was proposed to deal with the problem. The definite of simplified decision table is first introduced, and many duplicate objects are deleted in decision table. Then the subsets of discernibility matrix is constructed with respect to different decision classes,which effectively avoids the imbalance of objects and compresses the storage space of algorithm, and adopting the heuristic search strate- gy with backward greedy to calculate the relative minimal attribute reduction. Some useful decion rules based on reliabi- lity are extracted, what＇s more, the reliability is dynamically given, and the algorithm has good adaptability. Finally, ex- ample analysis and experiential results show that the proposed algorithm can exact effective decision rules from incon- sistent decision table.
出处
《计算机科学》
CSCD
北大核心
2013年第6期215-218,共4页
Computer Science
基金
国家重点基础研究发展计划项目(973计划)(2009CB522701)
国家自然科学基金项目(60963008,60875029)
科技部创新方法专项项目(2010IM020900)资助
关键词
粗糙集理论
不一致决策表
属性约简
规则获取
Rough set theory, Inconsistent decision table, Attribute reduction,Rule extraction
