不完备信息条件下基于Roustida改进算法的诊断规则提取

诊断决策过程本质上为信息的处理过程。由于信息结构的复杂性和采集的局限性使得获取的信息存在缺失、模糊、冗余等不完备现象,从而影响诊断的准确性。为此,对条件属性冗余、部分数据值缺失情形下,如何提高被诊断信息的完备性开展讨论,试图通过问题聚类探寻诊断决策所需的隐含规则,提出信息补齐与属性约简的知识挖掘方法：首先,针对Roustida算法在缺失值处理时存在的局限性进行改进,扩充其在工程实践中的适用范围,使缺损信息趋于完整；然后,利用遗传算法和广义诊断规则推理实现条件属性约简和规则凝练；最后,以质量问题诊断为对象进行了案例研究,测试样本诊断结果覆盖度 ,验证了不完备信息条件下该方法可以实现以相对较简方式表达问题与情境信息之间的关联关系,挖掘问题发生的隐含规律。

Diagnosis rule extraction based on improved Roustida algorithm under incomplete information

The decision-making process is essentially the information processing. Due to the complexity of information structures and the limitation of information collection, there are often incomplete information phenomena such as missing, fuzzy, redundant, etc., which affects the accuracy of diagnosis. For this reason, this article discusses how to improve the completeness of the diagnosed information when the conditional attributes are redundant and some data values are missing, attempts to obtain the implicit rules required for diagnosis through problem clustering, and propose the rule extraction method of information complement and attribute reduction. Firstly, the limitations of Roustida algorithm in processing missing values are improved, so as to expand the scope of its application in complex engineering practice and to make the missing information complete. Then, the conditional attribute reduction and rule extraction are carried out by using the genetic algorithm and the generalized diagnostic rule reasoning respectively. Finally, A case study was conducted on the diagnosis of quality problems: the coverage of the diagnostic results of the test samples was 87.5%. Under the condition of incomplete information, the proposed method can express the relationship between the problem and the situational information in a relatively simple way, and mine the hidden knowledge or rules of the problem occurrence.