Interactive Genetic Algorithm Based On Gene Classification Research And Application

Interactive genetic algorithm (IGA) is a kind of genetic algorithm which obtains individuals'fitness based on people's subjective evaluation. It combines human intelligence assessment and evolutionary computation, breaks through the explicit performance index restrictions by establishing optimal system, greatly extends the application area of evolutionary computation. However, interactive genetic algorithm itself is limited by users easy to be tired, compared with computer, so the population size is not too large and cannot represent vastly different individual objects effectively. At the same time, the man-machine interface and the system output characteristics restrict the features of the algorithm and range of prospects in real applications.To solve the problem of slow convergence speed of large-scale population and user's fatigue owing to a long period of interaction, an idea of gene-level classification is introduced into interactive evolutionary computation, and the algorithm based on genetic-level classification is proposed. The main idea of this algorithm is that according to the individual characteristics of chromosomes pre-defines hierarchy of categories of individual genes in the initial coded stage, divides the global search space into different local space on the basis of their effective range. In the evolutionary process, it can quickly narrow the global search space to the local search space through property levels, accelerates the convergence speed.Meanwhile, because there is poor local search efficiency in the interactive genetic algorithm, the key solution through people's participation in choosing display type in the local gene segment, speeds up the local search efficiency. To solve the problem of slow convergence rate and easy to fall into the shortcomings of local optimum, some improved strategy inheriting better genes to the next generation by targeting part of the excellent ones is made. Dispersing solution space in initialization stage and uniformly initialing population as far as possible increases the possibility of access to the global optimal solution. The improved algorithm can effectively reduce the invalid crossover operation. Convergence rate, global and local search capabilities have been greatly improved. The algorithm is detailed introduced in this paper, and applied to the fashion design. It is better than direct manipulation of interactive genetic algorithm both in the average convergence algebra and probability of convergence to the optimal solution.Finally, conclusions of the thesis are made. The deficiencies and suggestions are proposed for further research.