Multi-stage Decision Model For Aggregating Linguistic Information And Its Application

The multi-stage decision-making is the extension and enlargement of static decision-making problem in the time dimension, which is commonly utilized in some decision-making practices with a long range or several states, such as construction project evaluation, economic development appraisal and venture investment project evaluation. As the performance of the decision-making objective behaves a certain dynamic evolution, many evaluation outcomes in several stages should be considered to obtain the comprehensive decision-making result. Because of the diversity of decision environment, the complexity of decision objective and experts’ limited rationality, the performance of alternatives, especially under the qualitative indexes, couldn’t be described with accurate numbers in most time. Consequently, linguistic information is often employed in various decision-making problems, for which decision-maker’s thinking habit and logic can be reflected properly with a good effect.The multi-stage decision-making problem often takes a long time, in which may contain much appraisal information from various sources. There are interrelationship and differences among these information necessarily.The related research about aggregating linguistic information only stays in the exploration state, which have not been mature enough to solve realistic questions with a good effect. Consequently, the combination area of linguistic decision-making and multi-stage decision-making is explored for aggregating linguistic information in several stages, which can supply theory proof and technical support for the multi-stage decision-making with linguistic information.According to the research mentality of from shallower to the deeper and from easy to difficult, the related research of this thesis can be divided into the following six parts, which reflects the six different decision-making situations.(1) The aggregating method for linguistic information in multi-stage decision-making problem is proposed to deal with Orness parameter and decision-making information, which can be utilized in the situation with expert’s preference about stages. Specifically, an analytical framework with TOPSIS(Technique for Order Preference by Similarity to an Ideal Solution) is designed, in which the dynamic features of alternative preference can be studied. Orness parameter is employed to describe the expert’s preference on the stage weights and an objective programming model is established concerning the stable development of stages, in which the stage weight can be obtained. Consequently, the comprehensive performance covers all stages can be aggregated with obtained stage weight. Additionally, the estimate model for the closeness range of the alternative performance is designed to reduce the information loss during the process of aggregating the uncertain linguistic information. Finally, the range analysis of Orness parameter under the constraint of alternatives’ order is conducted to analyze the influence of parameters’ variation on the sequence of alternatives, which can assist experts to select proper Orness parameter according to his preference.(2) The multi-stage stochastic linguistic decision-making method referring to the prospect theory is developed, which can be used in the situation with few experts’ preference on the stages. Specifically, the expert’s risk attitude in the dynamic decision-making process is explored to describe the developing feature of alternatives in stages with prospect theory. What’s more important, a novel method to determine the dynamic referring point is proposed. An objective programming model is established for seeking the maximal prospect values of alternatives, in which the criteria weights in all the stages can be determined. Furthermore, the comprehensive prospect values of alternatives in several stages can be calculated with the dynamic feature of referring points. Besides, another objective programming is designed to estimate the range of comprehensive performances of alternatives, which can be used to study the influence of decision-making risk on evaluation result.(3)The multi-stage decision problem under dual kinds of linguistic information is developed, which contains two kinds of linguistic information as decision support information and expert judgement. Detailly, the relationship and transforming rules of decision support information is studied and the fusion method of dual linguistic information is proposed. Additionally, the dynamic evaluation of linguistic information is explored and a group of programming models is set up to determine the time weight and criteria weight in each stage hierarchically, which can confirm the divergence of dual information can reach its minimum. Furthermore, the expert’s preference on two sorts of information is studied, which can be employed to combine the decision support information and expert’s judgment result. Through aggregating the appraisal performances in all the stages, the alternative sequence can be determined.(4) A novel multi-stage group decision method under dual linguistic information is proposed, which concerns the mutual revision of experts’ preferences and can be employed in the situation contain both decision support information and experts’ preference information. Specifically, the matrix of experts’ comprehensive preference is designed, which can reflect the overall effect of dual linguistic information. Additionally, the deviation degree between expert’s preference matrix and comprehensive preference matrix is calculated and an objective programming model can be established to find the weak-effective experts in the group whose judgment cannot meet the threshold requirement. Additionally, the experts’ preference can be described as a multidimensional space vector and the revision direction and optimal moving length of weak-effective experts’ preferences can be confirmed. According to the revised experts’ preferences, the relationship between group preference deviation and stage weight is studied. Furthermore, a programming model is designed to determine the stage weights by minimizing the differences between adjacent states. Finally, the overall stage performance of alternatives can be aggregated through the stage weights, which can be used to select the optimal scheme.(5) A novel clustering method dealing with dual linguistic information fusion is proposed to solve the conflict of clustering results caused by different sorts of information, which can be utilized in the multi-stage decision-making with more than 10 experts. Specifically, the experts’ similarities are analyzed according to preference information, which can be employed to obtain the prior group classification preference. With the similarity vector of decision support information, the consistency and inconsistency indexes are introduced to present the extent of uniformity and difference of group clustering results from dual-dimension calculation. With the objective of minimizing the conflicts of group clustering results under dual information, a programming model is constructed to calculate the attribute weights. Furthermore, the fusion degree of dual linguistic information is calculated and the within class and inter class weights are designed, which can aggregate the group preference under dual information in each stage. Finally, the relationship between information fusion degree and stage weight is studied and the above clustering method can be used in the multi-stage decision-making problem. Consequently, the alternative performance in each stage can be aggregated and the most suitable scheme can be confirmed.(6) The case about the construction performance of social credit system in the Yangtze River Delta region is evaluated in five stages with linguistic information. Specifically, the framework and basic components of social credit system in China is studied, which can be used to design its evaluation indicator system. Besides, the current situations of provinces Yangtze River Delta region in the period of “Eleventh Five”is investigated, which can supply the decision referring support to the experts. Additionally, the construction performance of social credit system in the Yangtze River Delta region under decision support information and dual linguistic information is evaluated, which can show the application steps of the above models and methods and verify the its scientificity and practicality.