关键词: 最邻近法, 注塑产品, 粗糙集, 事例检索, 模拟退火算法, 事例推理

Abstract: According to the characteristics of injection products, a K-Nearest Neighbors (K-NN) case retrieval strategy was proposed based on rough set and simulated annealing algorithm. Through use of rough set, case attributes were reduced and the importance sequence for attributes was obtained, which was employed to build case hierarchy clustering model and allocate attributes weights. With reduction results, case base was organized as a hierarchical structure therefore the case retrieval space was reduced and the retrieval efficiency was improved. By using of simulated annealing algorithm, attribute weights of hierarchy clustering model were optimized in terms of weights allocation quotas, which was used to improve the case retrieval efficiency and quality. The K-NN case retrieval strategy has been applied in injection mould design system which indicated that the strategy could dramatically improved overall case retrieval efficiency and quality in case-based reasoning system.

Key words: K-Nearest Neighbors, injection product, rough set, case retrieval, simulated annealing algorithm, case-based reasoning