Research On IEC-based Decision-Making Methods Of Hybrid Multiobjective Decision-Making Problems With Tacit Objectives

In the management and decision-making field, there exit many Hybrid Multi-objective Decision-making Problems with Tacit Objectives (HMDMPTO) which are important problems in the research area of Tacit Objective Decision-Making Problems (TODMP). The HMDMPTO problems have three marked features: the objectives in HMDMPTO problems are consisted of some objectives which are hard to be defined explicitly and other objectives which are able to be defined explicitly, the problems also have a very large number of decision solutions, even infinitely many; the preference structure of decision-maker is of uncertainty, and it needs to be confirmed gradually in the interactive process of problem solving. For examples, the ergonomic chair design problem, the manufacturing plant layout problem and the travel itinerary planning problem. It is the intrinsical requirement of the TODMP research to develop the appropriate method for the HMDMPTO problems, which is of great theoretical significance and application value. The marked features of the HMDMPTO problems make it hard to very complex to solve the kind of decision-making problems and unable to apply the traditional method such as weighted sum methods, utility function methods and compromise methods etc. The problem-solving process of the HMDMPTO problems requires interactive mechanism so that the preference structure can be obtained and conceived gradually. In recent years, with the development of intelligent optimization algorithms and artificial intelligence technologies, the Interactive Evolutionary Computation (IEC) methods with human-computer interaction mechanisms has become a strong advantageous method suitable for solving the HMDMPTO problems, especially its interactive mechanism and merits for solving complex problems inherited from the Evolutionary Computation methods.This dissertation analyses the difficulties of solving the HMDMPTO problems from the angle of introducing the features of the HMDMPTO problems, then studies some intelligent decision-making methods of the HMDMPTO problems on the basis of IEC and the decision-making supportive model for the HMDMPTO problems, in combination with multi-agent intelligent and artificial immune computation techniques. The main achievements of this dissertation are as follows:(1) In the view of the lack of decision-making supportive model for the HMDMPTO problems in current research, in this dissertation a novel decision-making supportive model for the HMDMPTO problems is proposed by melting the interactive decision-making mechanism of IEC and the Multiobjective Optimization Evolutionary Algorithms (MOEAs), and the concrete problem-solving process of the instantial method constructed by melting the IEC and the Archived MOEA method under the structure of the proposed model to illustrate the operability of the model.(2) In the view of evolutionary efficiency problem of the IEC-based method for the HMDMPTO problems, this dissertation proposed a novel Roulette Inversion Operator (RIO) of higher stochastic search ability, and the RIO operator is integrated into the design process of the methods for the HMDMPTO problems in order to improve the evolutionary efficiency from the angle of the algorithm mechanism. Theoretical analysis proves the RIO operator can conquer the intrinsical shortcoming of John Holland's Inversion Operator (HIO) when used for real coded stochastic algorithms. The experiments of several benchmark functions also show the superiority of the RIO operator. Then according to the features of the HMDMPTO problems, an interactive multi-agent multi-objective optimization evolutionary algorithm (IMAMOEA) is developed by melting RIO operator with the multi-agent computation technique to define the Agents, Living Environment for Agents and behavior rules of the Agents such as dispersion, mutation, competitive death, rebirth and self-learning. The IMAMOEA takes good use of human intelligence and the characters of agent-based computation technique to let the users only need to select the best and the worst individuals of the candidate population which can effectively shorten the evaluation time and therefore reduce the user's evaluation burden. The experiment for the manufacturing plant layout problem shows the validity of the proposed IMAMOEA algorithm.(3) The diversity-missing problem of feasible solution population in the IEC-based decision-making process of the HMDMPTO problems. In this dissertation, an adaptive strategy for maintaining population diversity is designed by integrating the population entropy sampling method and a adaptive mutation operator. In order to examine the validity of the designed strategy, a small population genetic algorithm integrating the diversity-maintaining strategy is developed, and the numerical experiments show the good performance of the genetic algorithm and validity of the strategy. Furthermore, an interactive multi-objective evolutionary algorithm (IAMEA) with the application of the diversity maintaining strategy is proposed. The experiment for the manufacturing plant layout problem shows the validity of the proposed IAMEA algorithm.(4) In the view of evolutionary efficiency problem of the IEC-based method for the HMDMPTO problems, in this dissertation, an interactive immune multi-objective evolutionary algorithm (IMMEA) is designed by introducing the artificial immune computation technique into the IEC research area, furthermore, another immune-based interactive multi-agent multi-objective evolutionary algorithm (IIMOEA) is proposed by fusing the immune cloning strategy into the multi-agent computation technique to define the immune Agents (antibodies), Living Environment for the immune Agents and the behavior rules for the immune Agents such as antibody polyclonal behavior, antibody monoclonal behavior, antibody death behavior, antibody rebirth behavior and antibody self-learning behavior to support the decision-making of the HMDMPTO problems. In both IMMEA and IIMOEA, decision-makers are asked only to select the best and worst solutions of the current candidate individuals which makes the evaluation process easy and therefore reduces the decision-makers' evaluation burden. The experiments for the apparel-purchasing recommendation problem show that both the IMMEA and IIMOEA are valid and better than the traditional sequential interactive genetic algorithm.(5) A prototype of decision support system for the HMDMPTO problems is studied, and its modules are discussed. And the apparel-purchasing recommendation problem as an instance of the HMDMPTO problems is studied. The coding and solving ideal of the apparel-purchasing recommendation problem is analyzed and discussed. The implement flow of apparel-purchasing recommendation system based on IEC is studied. The functions of the developed system are discussed.The HMDMPTO problems are a class of complex decision-making problems, and these problems are pervasive in our real life. From the perspective of multi-objective decision-making, this dissertation firstly constructed a novel decision-making supportive model for the HMDMPTO problems, then this dissertation studies and proposes several new intelligent IEC-based decision-making optimization methods for the HMDMPTO problems in order to improve the HMDMPTO's decision-making efficiency. The results of these studies enrich the contents in the TODMP research field, and can provide methodological guidance and technical support for the actual decision-making.