基于CBR系统事例检索算法的研究

Case Retrieval Algorithms in Case-Based Reasoning System

事例检索是基于事例推理CBR系统的中心环节,检索速度和精度关系着整个系统的质量。根据应用的需要先后分析了基于数据库、ID3决策树和k d树的事例检索算法,并在标准的k d树算法的基础上引入聚类的概念,将原始的事例库组织成聚合中心库和类库二级结构,提出了改进的k d树事例检索算法,理论分析和对比实验证实了改进的k d树事例检索算法在事例检索速度上逼近ID3决策树检索算法,在检索精度上远远超过了ID3决策树检索算法和标准的k d树检索算法,其性能开销比是四种算法中最为理想的。

Case retrieval is the focus of the CaseBased Reasoning system, and it determines the speed and accuracy of the whole system. First, We compare the algorithms based on SQL, ID3 and standard kd tree. Then, we introduce clustering into standard kd tree, so we can divide case base into a twolevel structure, clusteringcenter base and class base, to establish an advanced kd tree. The analyses and experiments have proved that the retrieving speed of our approach nears ID3s, and its accuracy is much better than others.