体现分布特征的混合信息随机转化聚合求解方法及应用

Method and application of random transformation and integrated aggregation of heterogeneous information based on distribution characteristics

随着决策环境的日趋复杂,多类型混合信息共存的现象越来越普遍.对此,考虑不同类型决策信息的分布特征,细化探讨以分段随机抽样方式将混合信息转化为随机数的方法,能够尽可能提升转化后数据与原始信息分布结构的一致性.通过大规模仿真归纳总结了不同情形下混合信息随机转化过程中最佳分段子区间的求解方法和结论,可为实际中分段抽样区间的确定提供参考.进一步,融合随机聚合求解与混合信息的随机转化过程,得到优胜度概率矩阵,并据此推导出排序位概率矩阵和可能性排序结论,从而实现决策结论由“绝对形式”向“相对形式”的柔性转化.该结论形式更加吻合原始决策信息的多样化、不确定性及模糊性特征,更具可解释性和可接受性.最后,通过对比分析验证所提出方法的有效性,并以省域发展潜力综合预判为背景构建应用算例,对所提出方法的特征及应用前景进行说明.该研究进一步丰富了混合信息的转化途径,可为多类型信息共存的复杂决策问题提供方法支撑.

With the complexity of decision environments,the coexistence of heterogeneous decision information is more and more common.The paper explores the transformation method to unify heterogeneous information into a set of random values by segmenting the data range into several subintervals based on its distribution characteristics.This method can improve the consistency of distribution structure between the transformed values and initial information as much as possible.Through large-scale simulation,the solution method and conclusion of the optimal sub-segment interval in the random transformation process under different situations are summarized,which can provide references to the determination of the sub-segment sampling interval in practice.Further,the integrated aggregation method is fused with the random transformation process to obtain the superiority-based probability matrix,from which the ranking-based probability matrix and the possibility ranking result are deduced.This type of ranking result achieves the flexible transformation of decision conclusion from“absolute form”to“relative form”,which is more consistent with the diversity,uncertainty and fuzziness of the original decision information and is more interpretable and acceptable.Finally,the effectiveness of the proposed method is verified by comparison analysis.An application example,regarding comprehensive prediction of provincial development potential,is constructed to illustrate the characteristics and application prospect of this method.This study further enriches the transformation approach of heterogeneous information and provides method support for complex decision-making problems with multiple types of information co-existing.