关键词: 案例推理, 灰色关联相似度, 聚类, 最近邻, 案例对齐

Abstract: This paper proposes aggregation retrieval for fault cases to resolve limited application of traditional nearest neighbor (NN) in accurate attribute, direct ratio relation between time consumption and case base scale, and no correlation of retrieved case and expected results. By introducing zone as representation mean of uncertain data, the paper creates the similar computation models of five kinds of attributes and produces whole similar model of cases based on grey-relational theory. Further, it gives an improved retrieval algorithm by organizing index according to aggregation analysis, and evaluates the retrieved results by rank correlation alignment measure. Lastly, an example explains the feasibility of aggregation algorithm, two groups of experiments show the proposed algorithm effective and innovative in performance.

Key words: case-based reasoning, grey-relational similarity, aggregation, nearest neighbor, case alignment