Abstract: Aiming at the basic contradiction between index accuracy and speed in case retrieval,as well as the problem of retrieval effectiveness caused by the increase of instances,the improvement and fusion research based on correlation function and D-HS index was carried out.The concept of fuzzy correlation function was proposed and the corresponding formula was deduced by combining fuzzy number and correlation function.At the same time,combined with Analytic Hierarchy Process (AHP),the method named unidimensional correlation function combination similarity measurement model was established.The product attribute values were transformed non-linearly based on the Sigmoid function,which improved the ability of data hierarchy division and resolves the problem of low retrieval accuracy caused by non-uniform distribution of attribute values in the traditional D-HS index.The validity and feasibility of the research were verified by integrating the above improved methods and taking the retrieval of vacuum pumps as an example.

Key words: case retrieval, retrieval utility problem, correlation function, D-HS index, fuzzy numbers

摘要： 针对实例检索中普遍存在的索引精度和索引速度间的基本矛盾问题,以及由于实例数目增多导致的检索效用问题,进行了基于关联函数和D-HS索引的改进与融合研究。首先,结合模糊数和关联函数,提出模糊关联函数的概念并推导出相应计算公式,同时结合层次分析法建立单维关联函数组合相似性度量模型;其次,基于Sigmoid函数对产品属性值进行非线性变换,提高数据划分层级的能力,解决传统D-HS索引中因属性值的非均匀分布导致的检索准确率低下问题;最后,融合上述改进方法,并以真空泵的检索选型为例,验证了研究的有效性和可行性。

关键词: 实例检索, 检索效用问题, 关联函数, D-HS索引, 模糊数