Research On QGA-based Methods Of Multi-objective Decision-making Issue

For persons who engage in administration and decision-making field,multi-objective decision-making problem is a common one, which can divide intotwo types, according to the existence of fixed targets or not. For problems with afixed target, precise solution can be reached with mathematical methods, based ontarget modeling. However, most problems in real applications does not have fixedtargets, which is known as semi-structured or non-structured problems, when facingthese problems, it may be difficult to make a decision due to the reason that targetor principle of decision can not be expressed in forms of quantities andstructures,these so-called recessive multi-objective problems have common featureas follows, the first one is that the target of decision can not be completely whollydescribed in form of quantity and structure, the second is that the decision is alteredwith preferences of decision maker, and the last one that the problem solving oftenends with failures due to the nature of ‘NP-Complete’, which make traditionalmethod fail in solving them. When dealing with these problems, we must first focuson tasks such as describing problem, coping with the changing preference, andimproving searching efficiency. So, with the assist of techniques such as artificialintelligence, therefore, the aids of how to solve these common recessivemulti-objective problems through effective model and method in the aspect ofdecision-making theory, is still a big challenge, but presents value in both theoryand application.Interactive Evolutionary Computation (IEC) is a kind of evolutionarycomputing method of which the fitness is assessed and finished by human, can gowell with recessive multi-objective decision-making issues. In this dissertation,Quantum Genetic Algorithm has been introduced into IEC, to solve multi-objectivedecision-making problems without fixed targets, with theories of multi-objectivedecision-making and technical analysis.First of all, this paper introduces multi-objective decision-making issues fromfour aspects, including problem developing, theoretical model, solution method andrelevant applications, and divide them into two types according to whether thetarget function is fixed or not. For models with fixed targets, Genetic Algorithm,which is a common solution method, along with the fundamental thought, characteristic and basic operation of it. The followed part is on Quantum GeneticAlgorithm, which extract much focus in recent years, basing on knowledge ofessence and procedure of this method, we conclude that it is suitable for problemswith fixed targets. We close this part of work with classic TSP problem, accordingto experimental comparison of traditional Genetic Algorithm and Quantum GeneticAlgorithm, the advantage of latter one in applications with smaller population scaleand less generation is verified.For multi-objective issues without fixed targets, this dissertation presentsevolutionary model in first, followed by the introduction of the interactive method,which is key to these problems, and choose planning of journey as an individualcase for analyzing. This case is extended from TSP problem and does not have afixed target, these two features, accord well with the typing method tomulti-objective issues of this dissertation, which is significant in comparison. Inanalyzing works on this case, we combine Quantum Genetic Algorithm withinteractive mechanism, and compare the result to traditional interactive GeneticAlgorithm statistically, therefore we conclude that the Quantum Genetic Algorithmis an effective method to solve this type of problems.