WDRR:一种带权重的多维案例推理算法

WDRR: A Weighted Multi-dimensional Case-based Reasoning Algorithm

案例推理技术已经成为故障诊断、管理辅助决策、专家系统等实现的重要手段.现有的案例推理算法针对海量案例集时,普遍存在检索效率不高问题.设计了一种带权重的多维案例推理算法(Weighted DimensionReduction and R-tree,WDRR),该算法结合案例的多维特征权重,将多维案例降维成二维案例点,并在此基础上建立R树空间索引;案例检索时首先借助R树索引,确定案例的二维点所在,再结合二次权重和K近邻(KNN)算法进行精确过滤,根据相似度阈值输出案例推理的结果,并完成案例学习和索引修正.实验证明该方法针对海量案例集的检索效率和准确率都有较大的提升.

Case-Based reasoning technology has become the important method to realize fault diagnosis, aid decision, expert system etc. Generally ,there exists low retrieval efficient problem when the current case-based reasoning algorithm is used in large scale case set. This paper designs a weighted multi-dimensional case-based reasoning algorithm（Weighted DimensionReduction and R-tree, WDRR ）, which reduces dimensionality from a multi-dimensional case into a two-dimensional case point, combining with the multi-di- mensional feature weights of case, on the basis, then establishes R-tree spatial index. When retrieving cases, firstly it uses R-tree index to determine the two-dimensional point of the case, then combined with second weight and K Nearest Neighbor （KNN ） algorithm to filters cases precisely ~ finally according to the similarity threshold to output the result of case reasoning and accomplish the case study and the index correction. Experimental results show that this method can achieve high performance on both retrieval efficiency and ac- curacy for massive case set.