带Rough算子的决策规则及数据挖掘中的软计算

DECISION RULES WITH ROUGH OPERATOR AND SOFT COMPUTING OF DATA MINING

文中讨论决策规则及其与演绎推理中的假言推理规则之间的关系．通过数据挖掘中的软计算使决策表中的属性简化和属性值区间化，从而找到一种具有广泛表达能力的数据隐含格式，从中选择有代表性的、并删去冗余或过剩的规则，并保持决策表的原有用途和原有性能．我们通过开发一个中医诊疗专家系统的实例说明了这种软计算的过程，并分别用基于统计或专家经验方法计算带可信度因子的产生式规则和基于Ｒｏｕｇｈ集方法计算带Ｒｏｕｇｈ算子的决策规则两种不同方法开发同一个系统，可以看出用Ｒｏｕｇｈ集方法计算带Ｒｏｕｇｈ算子的决策规则较之１９８７年用统计或专家经验方法计算带可信度因子的产生式规则更加理论化和实用化．经临床试用已初步显示出其优越性．

Decision rules based on rough sets and the relationship between the decision rule with modus ponens rule in the deductive reasoning are discussed in this paper. The attributes in decision table are reduced, and the values of attributes in decision table are made an interval via the soft computing of data mining, whereby finding a reducible implied pattern of the data, and choosing the representative, while deleting those redundant or superfluous decision rules in the decision table. And it can keep the original properties and functions of the decision table. The procedure of soft computing by developing an expert system for diagnosis and treatment in acupuncture and moxibustion in Chinese traditional medicine is also described. The system developed based on rough set approach is more theoretical and common applicable than that developed based on statistics and expert experience in 1987. It shows preliminary superiority via the clinical tests in the traditional Chinese medical science.