多关系混合不确定信息的随机模拟聚合方法及应用

针对综合评价中混合不确定信息共存、评价信息残缺和评价信息间非独立的评价问题，提出一种多关系混合不确定信息融合集成框架及其求解方法.首先，将混合不确定信息进行分类整合，梳理信息(子)流间的相关关系，构建多关系混合不确定信息集成框架，通过网络分析法(ANP)求解信息(子)流的信息权；其次，将混合信息转化随机数并模拟迭代，充分挖掘信息价值；最后，将被评价对象每次仿真迭代后的结果进行两两比较，获取优胜度矩阵，并通过回归树的方法得到体现概率特征的可能性排序.对混合残缺信息、多关系信息的融合问题进行了探索，降低了评价过程对数据完整性的要求，考虑了信息重叠和信息冗余，得到的相对评价结论对实际问题更具解释力.

Stochastic Simulation Integrated Method for Multi-relational Blended Uncertain Information and Its Applications

In view of the problems of the coexistence of multi-source information， incomplete information， and non-independent evaluation questions of evaluation information， an integrated framework for multi-relational blended information and an information aggregation method are proposed. Firstly， the blended uncertain information is classified and integrated， the correlations are explored among information(sub)streams， and the integrated framework for multi-relational blended uncertain information is constructed. Secondly， the blended uncertain information is transformed into random numbers by simulation iterations to fully tap the value of information. Then， through the relationship graph of information(sub)flows， the information weight of information(sub)flows is solved by using the analytic network process(ANP). Finally， by comparing the results of each simulation iteration of the evaluated object， the pairwise priority matrix is obtained， and the probability ranking reflecting the probability characteristics is obtained by the regression tree method. The fusion of blended incomplete information and multi-relational information are explored by constructing the integrated framework for multi-relational blended uncertain information. The requirements for data integrity in the evaluation process are reduced， and information overlapping and information redundancy are considered. The relative conclusion is more explanatory for the practical problems.