基于粒计算的多属性群决策分析

Multi-attribute Group Decision Making Based on Granular Computing

通过仿生人脑智慧性思维,提出了基于粒计算的多属性群决策求解思路。首先给出了粒计算结构模型的数学刻画,而后为细致描述单个决策人对决策问题在粒层间的往返思考,给出了相对意义下的粒信息熵度量,并通过定义"粒划分匹配率"和"粒划分覆盖率",实现基于相似关系阈值进行单个决策人独立思考下的粒层搜索寻优计算。最终寻优结果体现于该决策人理解思维下的特征属性集权重向量值上。最后基于各个决策人思考粒层下的权重向量值,通过非线性优化模型的求解,在粒层间予以再次平衡寻优求解,获得决策群体一致认可的特征属性集最优权重向量值,实现对多属性群决策问题的一体化合理求解。仿真实例验证了本文方法的可行性和优越性。

By simulating the process of human thinking,a solving method for multi-attribute group decision making problem based on granular computing was proposed.Firstly,mathematical description of granular computing structure model was discussed.Secondly,in order to give a fine description to single decision maker’s thought at different granular layers,granular information entropy measurement under relative meaning was given.Through definitions of matching rate and coverage rate under granular divisions,the optimization of granular layer based on similar threshold was carried out,and the final optimization result,which indicates single decision maker’s decision thought,was exhibited by weight values of feature attributes.Thirdly,based on values of weight vectors which indicate decision makers’ decision thought upon different granular layers,the nonlinear optimization model was built up to rebalance all decision makers’ opinions at different granular layers and get the final optimal weight values of feature attributes,which can be recognized by all decision makers.Thus,an integrated and qualitative problem solving environment for multi-attribute group decision making was reached.Lastly,a graduate student admission interview assessment example was given to prove its feasibility and superiority.