实例检索中基于混合属性距离的相似性度量

Similarity Measurement for Retrieval Based on Hybrid Attribute Distance

摘要:

提出了一种基于混合属性距离的相似性度量方法.利用各个属性间的距离及其组合权重求得实例间的总体距离，再用实例间的总体距离来刻画其相似程度.基于区间数的距离计算公式和模糊集合理论，给出了属性值为模糊数、模糊区间数时的距离公式，并改进了属性值为隶属度函数时的距离公式.同时考虑了属性权重问题，提出了一种基于距离离差信息的客观赋权方法，将主观权重和客观权重加以组合，以组合权重来计算实例的全局相似度.以阀门的概念设计为例，验证了该方法在实例检索中的的可行性和合理性.

Abstract:

A new similarity measure method based on hybrid attribute distance was introduced. Overall distance between two cases was determined by distances between attributes and synthesis weights. First, a degree of similarity was characterized by overall distance. Based on the distance formula of interval numbers and the theory of fuzzy set, a distance formula of fuzzy numbers and fuzzy interval numbers was given and the distance formula of membership function was improved. Meanwhile, the deviation information of distance values was used to calculate objective weights regarding weights assignment, then objective weights and subjective weights were integrated into synthesis weights to calculate the overall similarity of the cases. Lastly, the method was applied to the conceptual design of valve. The effectiveness and feasibility of the method in case retrieval were demonstrated.