一种改进的不完备熵属性约简在装备故障诊断中应用

针对经典的粗糙集方法在处理不完备测试数据时,将完备化和属性约简先后进行易导致完备化数据产生偏差而影响属性约简准确性的问题,给出一种基于不完备熵的属性约简算法.通过对属性重要度概念在粗糙熵和不完备熵定义上进行拓展,立足原始测试数据直接计算各条件属性的属性重要度,从而获取一个较优的属性约简.该方法不仅充分考虑了不完备数据与属性约简之间的联系,而且约简结果更加精确,最后在某型装备故障诊断中应用验证了其有效性和优越性.

An Improved Attribute Reduction Algorithm of Incomplete Entropy Application in Equipment Fault Diagnosis

In view of the problem that completion and attribute reduction lead to completion data devia- tion and influencing the accuracy of the attribute reduction in the classical rough set method, an improved attributes reduction algorithm of incomplete entropy is given. The concepts of attribute importance on the rough entropy and incomplete entropy are expanded, and the attribute importance degree of each condi- tion attribute is calculated based on the original test data directly to obtain a relatively optimal attribute reduction. This method not only fully consideres the contact between the attribute reduction and the in- complete data, but also makes the reduction results more accurate, and is was applied to some equipment fault diagnosis to verify the effectiveness and superiority.