基于决策矩阵的可信关联规则挖掘方法

摘要

Aiming at the rule extraction in the information system ,a kind of method based on the decision matrix for mining credible association rule was proposed .The threshold parameter fitting the particular condition equivalence class was produced automatically ,and other condition equivalence classes were filtered to generate the condition discernible function ,then all credi-ble association rules corresponding with this particular condition equivalence were obtained after transforming the condition dis -cernible function to its extreme minimal disjunctive form .Based on this ,the matching cases between the test sample and the credible association rule set were also considered for designing the decision matching strategy ,and this algorithm was applied to the UCI data set .The experimental results and the data analysis show that the algorithm has comparatively steady performance on classifying prediction .%针对信息系统的规则提取问题，提出一种基于决策矩阵的可信关联规则挖掘方法。依据各条件等价类自动设置阈值参数，过滤其余可分辨条件等价类生成条件决策分辨函数，将此分辨函数转化为等价的极小析取范式，得到蕴含于该条件等价类的全体可信关联规则；在此基础上进一步考虑待测样例与可信关联规则集的匹配情形设计决策匹配策略，将该方法与策略应用于UCI数据集。实验结果表明，该方法具有较稳定的分类预测性能。