混合缺失信息集结下的属性赋权方法  

Attribute weights-given method under aggregation of hybrid and deficient information

作　　者：王昱[1] 傅莉[1] 梁宵[1] 李勇[2] 

机构地区：[1]沈阳航空航天大学自动化学院,辽宁沈阳110136 [2]沈阳工业大学电气工程学院,辽宁沈阳110870

出　　处：《浙江大学学报（工学版）》2016年第9期1806-1814,共9页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(61503255);辽宁省教育厅科学研究一般资助项目(L2015412)

摘　　要：针对数据缺失的多类型评价信息混合集结问题,提出信息的一致性插补方法并建立插补优化模型,给出各类信息Mass函数统一转换方法.提出基于信息不完全折扣因子、Pignistic概率距离和顺序贴近度的专家权重计算方法,确定Mass函数修正因子.应用Dempster组合规则融合属性权重,形成一种基于缺失信息插补和证据理论的属性赋权方法.仿真算例表明,该方法对专家信息重要性的考量标准不仅取决于属性评价信息的数据相似度,而且增加了信息数值排列顺序的贴近性以及专家所提供信息的完整程度,提高了属性权重计算的合理性.仿真结果验证了该方法的有效性.An attribute weights-given method was proposed based on the interpolation of deficient informa-tion and evidence theory, aiming at the hybrid aggregation problem of multi-type evaluation information with deficient data. F irs t, an information consistency interpolation method was put forward; the interpola-tion optimization model was established; a unified conversion method of Mass function was given. Then, the expert weights calculation method was proposed to determine the amending factor of Mass function based on the incomplete information discount factor and the closeness degree with Pignistic probability dis-tance and sequence. Finally, Demspter rule was applied to combine the attribute weights. Simulation ex-ample shows that the consideration criterion to the importance of expert information not only rests on the data similarity of attribute evaluation information, but also the closeness of data numerical order and the completeness extent of information. Therefore, the rationality of attribute weights calculation is im-proved. The effectiveness of this method is verified by simulation results.

关 键 词：混合缺失信息 信息插补 证据理论 专家权重计算 属性赋权方法 

分 类 号：C934[经济管理—管理学]