中文摘要：

对于特征属性值存在缺失的CBR（Case-Based Reasoning）信息系统，使用经典粗糙集理论无法进行权值计算；同时，在CBR系统的实际应用中，计算属性的权值应避免出现结果为0的情况；首先分析了容差关系模型下使用条件熵进行权值计算的方法，然后根据属性值是否缺失分析各属性对整个系统的重要度，综合两种方法的优点，提出了一种新的权值计算方法，提高了权值计算的精确性和严谨性；将该权值计算方法应用于飞机故障诊断专家系统中，取得了满意的结果，验证了其可行性。

英文摘要：

Attribute weight can not be computed by making use of classical rough set, in the incomplete CBR system that some attribute value is unknown. Meanwhile, the situation that the result of computing weight is 0 should be avoided in actual application of CBR system. The method of computing weight is analyzed through using the conditional entropy in the tolerance relation. According to attribute value being known or not, significance of the attribute relative to each system is discussed. A new method that brings together advantage of that two methods is presented. The accuracy and rigor are promoted. An example of the fault diagnosis expert system of aircraft has been used to illuminate the validity of the new method. The result turns out to be very good.