基于案例-规则检索的特征阈值选择模型

Feature Threshold Selection Model Based on Case-Rule Retrieval

基于案例规则的信息检索模型因知识不确定性或推理不可靠性而存在匹配效率不高的问题,因此制定合理高效的匹配策略来增强信息检索的准确性显得非常重要。本文提出融合酉空间的概念,并通过归一化处理将异构的案例和规则知识在空间中统一表示,定义检索模型的效能解;其次利用酉空间奇异值分解定理得到检索模型的阈值向量,引入融合推理传递函数矩阵的能量泛函的基础上证明了阈值的有效性;再次采用特征阈值向量界定了知识源的元数据与决策目标解集之间的关联性;最后根据特征阈值向量制定融合检索策略并提出了采用特征阈值的案例-规则融合方法实现知识融合,使用基准数据验证本方法的有效性。

Information retrieval model based on case-rule exists low matching efficiency due to the uncertainty and unreliability of knowledge fusion system, so developing a reasonable and efficient matching method to enhance the accuracy of the retrieval model is important. Firstly, this paper proposes the Fusion Unitary Space and defines the Heterogeneous cases and rules in knowledge space by normalization processing, then defines the robust solution of fusion reasoning model. Secondly, we get the threshold vector of system by using the decomposition theorem of uni- tary space singular value, and proves the effectiveness of the threshold based on introducing the Pan Function matrix of fusion reasoning method. Thirdly, we determine the relevance between metadata of knowledge sources and target solution set of decision by using the feature threshold vector. Finally, we formulate fusion retrieval tactics based on feature threshold vectors, propose the case-rule fusion method to achieve knowledge fusion, and verify the effective- ness of the method by using the benchmark data.