Research On Decision Theory And Method From The Perspective Of Network Analysis

Decision Theory and Method(DTM, hereafter) is an important and key research field for the discipline of Management Science and Engineering(MSE, hereafter). DTM lays a solid foundation for the overall development of Management Science, and meanwhile it also becomes an efficient tool to enrich management practices and enhance the ability of solving the practical issues. Besides, Network Analysis(NA, hereafter), which enjoys a broad scope of applications, is not only one of the current cutting-edge research directions in many study fields, but also provides several new perspectives and angles for developing and even creating the methodology, especially in the field of MSE. This Ph.D. thesis aims to adopt the perspective of network analysis, namely utilize the creative power of interdisciplinary, in order to put forward new decision theories and develop new decision methods.The major research contents of this thesis include three refined innovation thoughts from network analysis and six new decision theories or methods to practice the above thoughts. Specifically, from the angle of “node” in the network analysis, this paper refines a perspective called “whole network analysis” and accordingly proposes two new decision methods to evaluate the node influence and the node similarity, respectively. From the angle of “edge” in the network analysis, this paper generalizes an idea called “establishing network” which emphases on grasping the relationships; as a result, it makes process in the comprehensive evaluation methods and could provide more reasonable evaluation and recommendation results. From the angle of “structure” in the network analysis, this thesis suggests a thinking process called “seeking for intermediate state” in order to make up for the deficiencies of the existing studies. For example, this thesis focuses on exploring the effect from network structure on its functions and understanding the formation and evolution of network structure.This thesis presents new decision-making methods or improves the existing achievements of existing researches based on the above mentioned three innovation thoughts, which reflect the major innovation achievements of this paper. In details,(1) Based on the perspective called “whole network analysis”, this thesis suggests defining a macro index of the whole network first and then pays attention to its change from the micro viewpoint. Following this perspective, this thesis presents a new model with two kinds of parameters, formally and mathematically, to find the influential authors and mine the valuable coauthors based on the idea that how the whole network’s prestige changes when some node is removed from it. Also, this thesis proposes a new decision method to measure and evaluate network node similarity by utilizing the concept of “mutual information” to measure the whole network’s information and calculating the information loss of seeing two nodes as the same. Besides, this thesis discusses the time-saving algorithm to solve these models, and further examine, by simulation analysis and test on real-world data, the accuracy of the newly proposed methods as well as compare them with these classic ones.(2) The idea of “establishing network” points out that the key of establishing network is to grasp the linkages and mine the reason of forming the edge in the network. As a result, it will transform a traditional decision-making problem that is seemingly not related to network analysis into a one with network structure. If so, a traditional decision-making problem could be solved by the cutting-edge network analysis so that it will extend the methods to deal with decision-making problem. In this thesis, we propose a new rating model by considering the purchasing networks formed by consumers and further present the new model’s properties and validate the model’s applicability. Through the designed comparison analysis and the empirical analysis, the model is showed to be better than the traditional ones such as Averaging and Analytic Hierarchy Process(AHP) in terms of consistency and its ability to deal with the missing data. Meanwhile, we make a new tree-network model by mining the connection characters of short message transmission and then solve the problem of mining the Short Message Services seed users. This study not only fills in the blank of the corresponding methods, but also provides a powerful tool for seeking the source of public opinions, for controlling public opinion propagation and for differentiated customer management from communication companies.(3) The thinking of “seeking for intermediate state” is adopted in this thesis to develop the existing studies related to the relationship among network structure, perception level and participants’ welfares as well as the methods related to network evolution. This thinking would help to relax the extreme hypothesis in the existing literatures, to supplement the contents not discussed in the existing literatures, and to provide more realistic assumptions. Since the research on network structure and its decision behavior on it is just starting, most of literatures gave theoretical results in harsh and extreme assumptions, so it is convenient for us to adopt the thinking process of “seeking the intermediate state” to perfect the existing conclusions. This thesis, based on a market consisting of one monopoly and several customers who are embedded in an economic network, relaxes the assumption that the decision-making participants have complete knowledge of network structure presenting in the existing literatures, and studies how the different perception levels about the network structure affect the two kinds of participants’ welfares, and then provide some good strategies for the monopoly to mine the information of the network structure. Moreover, this paper extends the existing literatures which can only analyses network evolution under the two extreme situations, namely the initial stage of network evolution and the one having experienced a long period of evolution to achieve a stable state, so that it puts forward a unified analysis framework for exploring and understanding the patterns of network evolution with different evolution periods. The framework is validated through simulation-based artificial data and on a real dataset from social media platform. The results demonstrate that this framework is capable to uncover the micro-foundations and the complex interactions of network formation with different evolution periods.The theoretical significance of this research is to introduce the cutting-edge network analysis perspectives into the classic decision theory and method, and to develop the decision theory and create the decision methods. As a result, this thesis will contribute to make a substantial progress for decision theory and methods. As to the network analysis, the viewpoint of structural analysis, connection and integral analysis and so on will enrich the existing decision theory and will broadly and profoundly change the existing analytic methods of decision theory. Since the network is ubiquitous, the practical significance of this research lies in a widely useful methodology for the discipline of management science, namely the refined ideas and developed methods in this thesis would be appropriate for proposing numerous decision-making methods and solving never-ending decision issues.