摘要
一阶规则空间中的θ-包容序是一个quasi-order序(自反的和传递的),采用通常的基于规则覆盖例子数目的评判标准,沿着这个quasi-order序挖掘一阶规则时,存在一阶规则的等价类问题,严重地降低了算法的搜索效率和规则的可读性。采用基于规则绑定数目的评判标准则可有效区分等价规则的优劣,正确地指导算法的搜索方向。
Theθ-subsumption induces a quasi-order(reflexive and transitive)on the space of first-order rules.There ex-ists the equivalence class problem in adopting the common evaluation criteria,based on the number of examples cov-ered by rules,along theθ-subsumption order to mine first-order rules.That will badly reduces the search performance of algorithm and the readability of rules.The evaluation criteria based on the number of bindings of rules can distinguish the quality of equivalence rules,and correctly guide the search direction of algorithm.
出处
《计算机工程与应用》
CSCD
北大核心
2003年第18期58-60,79,共4页
Computer Engineering and Applications
基金
国家自然科学基金(编号:60173014)
北京市自然科学基金(编号:4022003)资助
关键词
数据挖掘
一阶规则
遗传算法
data mining,first-order rules,genetic algorithm
