针对农机专业底盘参数属性多、参数类型差异大、属性不能无条件补偿等特点,传统实例推理中以属性加权评价为主的最近邻算法已不能满足实际需求.基于此,提出将级别高于关系的ELECTRE多属性决策方法应用到农机专业底盘的实例推理中,算法首先对农机底盘实例数据进行预处理,然后采用层次分析法确定属性权重,通过不断试验调整各属性不和谐性阈值后进行级别高于关系构造、挖掘,最终实现目标实例与数据库中实例的相似性评价,并以履带式稻麦收获机底盘实例进行验证.与经典的最近邻评价方法相比,算法除能返回总体相似度较高的模型之外,同时返回影响实例排序的关键参数,让用户拥有更多后续评价依据,总体效果较好.%Agricultural machinery professional chassis has many types of parameters and their difference between each other is so large that they cannot be fully compensated without any price.For these characteristics,the classical nearest neighbor algorithm which relies on weight properties in case-based reasoning is unacceptable in practice.Aiming at the situation above,the sorted method ELECTRE,belonging to multiple attribute decision,which was characterized by outranking relationship,was applied to the similarity assessment.In this algorithm,the first step was data preprocessing;then AHP was adopted to determine attributes' weights;after that harmony and disharmony thresholds were obtained by constantly experimenting adjustment;and then construction and mining of relationships were carried out;the ultimate goal was to accomplish similarity assessment between design case and database cases.After that,crawler of walking system and chassis prototype were used for verification of the algorithm.Compared with the nearest neighbor method,the algorithm not only returned the former models which had higher overall similarity,but also marked the key parameters that affected the order,allowing users to have more follow-up evaluation.To make the algorithm to be more practical,a prototype interface of CBR was developed and the input of critical parameters and output of results were demonstrated.The overall result was good.
展开▼
