基于关联系数标准差融合的置信规则库规则约简方法

Rule Reduction Approach to Belief Rule Base Using Correlation Coefficient and Standard Deviation Integrated Method

为解决置信规则库中由于前提属性数量过多引起的"组合爆炸"问题,前人已经提出了许多维度约简的规则约简方法.然而,这些方法在算法实现的难易程度、筛选属性的合理性、规则约简的数量和决策推理的准确性上难以都取得理想的效果,因此本文基于关联系数标准差法提出了新的规则约简方法,其核心思想是根据关联权重来约简置信规则库.本文还引入装甲装备体系作为分析实例,并重点在特殊方案和一般方案两种情形下分析基于关联系数标准差规则约简方法的适用性.实验结果表明,本文所提出的方法的约简结果具有较低效用偏差和较高相似度.

To solve the problem of combinatorial explosion in belief rule base （BRB） caused by too many anteced- ent attributes, many rule reduction methods that adopt dimensionality reduction techniques are proposed. However, these methods cannot simultaneously achieve the desired results, namely algorithm implementation, reasonability of filtering attributes, the number of reduced rule and the accuracy of reasoning. Hence, a new rule reduction method that uses correlation coefficient and standard deviation integrated method is proposed. The kernel method is based on the coefficient weight for reducing BRB. An armored system of systems also in- troduced to analyze the applicability of the rule reduction method using the correlation coefficient and standard deviation integrated method. The experimental results show that the BRB, which is reduced by the proposed method, has lower utility deviation and higher similarity.