论文部分内容阅读
针对置信规则推理作为系统控制器时的应用,提出一种置信K均值聚类算法用于置信规则库的结构识别。在构建好置信规则库的推理框架后,该算法通过对规则前项输入变量的历史数据进行挖掘,得到合理的置信规则库结构,提高推理与决策的精度。相对于传统专家知识确定置信规则库结构的方法,该算法的特点是:最优聚类与相邻评价等级之间的距离成正比,与人的认知能力相一致;最优聚类保证采样点以最小的距离靠近评价等级,也就是保证输入变量尽可能趋近置信规则前项。通过置信规则推理在集约生产计划中应用的案例分析验证了该算法的合理性和有效性。
Aiming at the application of confidence rule reasoning as a system controller, a confidence K-means clustering algorithm is proposed for structure recognition of belief rule bases. After constructing the reasoning framework of the confidence rule base, the algorithm digs the historical data of the input variables in the preceding paragraph of the rule to obtain a reasonable structure of the confidence rule base and improve the accuracy of the reasoning and decision making. Compared with the traditional expert knowledge to determine the structure of the confidence rule base, this algorithm has the following characteristics: the optimal cluster is proportional to the distance between adjacent evaluation levels and is consistent with human cognitive ability; the optimal cluster guarantees sampling Points close to the evaluation level with the minimum distance, that is, to ensure that the input variables as close as possible to the confidence rules of the preceding paragraph. The case analysis applied in intensive production planning through the rule-of-confidence test verifies the rationality and effectiveness of the algorithm.