【摘要】 以规则库为切入点,提出一个决策规则的批量增量更新算法。为所有新增对象建立一个等价类表,将原有规则库与等价类表进行高效匹配,根据新对象的不同匹配类型分别进行规则更新。该算法既适用于完备数据也适用于不完备数据,且只需访问2遍规则库就可以实现规则更新。理论分析和UCI数据上的比较实验结果都表明该方法优于传统方法。更多还原

【Abstract】 From the view of rule base,a batch incremental algorithm for rule induction is presented in this paper.An equivalence class table is built for all new objects.Original rule base is efficiently matched with the equivalence class table.The rule base is updated for the new objects based on the different match types.The algorithm can deal with both complete and incomplete data and accomplish the rule update through only accessing rule base twice.Theoretical analysis and comparison experimental results on UCI datasets demonstrate that the algorithm outperforms traditional algorithms.更多还原