The Study Of The Genetic Evolutionary Algorithm Based On Cloud Model

Since the popularity of computer, computational intelligence has been a hot research subject. According to the thought of bionics, some scholars put forward the simulated annealing algorithm, ant colony algorithm, genetic algorithm, particle swarm optimization algorithm, etc. Among them, the Genetic Algorithm has become an important branch in the field of computational intelligence, it has the characteristics of random global search, optimization and do not depend on the specific problem. At present, it has been widely used in control planning, combinatorial optimization, signal processing, image processing, etc. Cloud Model is a kind of uncertainty conversion model between qualitative and quantitative, which integrates the theory of probability and the fuzzy mathematics. Using the cloud model to express natural language, which provides a new research methods and ideas for optimization and uncertainty of the problem, it can describe the natural language’s fuzziness and randomness in a better way. Normal cloud is the most commonly used model which has the universal applicability and stable tendency. The current genetic algorithm is inefficient in convergence and search capability. Aiming at these problems, based on cloud model uncertainty relation of conversion capacity between qualitative and quantitative, the traditional genetic algorithm is improved to enhance its optimal performance and its convergence. Great deals of research works have been done in this thesis:First of all, we optimize the strategy of genetic algorithm, propose the cloud learning operator and combine cloud crossover operator with cloud mutation operator to control the genetic evolution of population. According to the characteristics of the genetic algorithm and cloud model, this paper puts forward each operator adaptive strategy, which mainly through using the three parameters of cloud model to control operator’s change. Algorithm strategy is that the cloud operator is applied to use large crossover, mutation and the learning operator in the early evolution to get a faster convergence speed. After the relative stableness of population, with less operator, we expand appropriately the searching range in order to obtain the optimal individual. We use the cloud model to ensure the stable tendency and randomness of genetic evolution, and to speed up the convergence of the genetic. And the classical function is used to test the algorithm and check efficiency of the algorithm.Secondly, we improve the algorithm’s optimization and enhance the memory of algorithm. In the evolution strategy of genetic algorithm, we put forward the threshold. When it meets the condition, the global optimization strategy is implemented. In the evolutionary process, the Y-conditional normal cloud generator instead of cross process is utilized, and the normal cloud generator instead of mutation is used to update the population. This method overcomes the non-memorization characteristic of the previous genetic algorithm. It can jump out of local optimal and realize the global optimization. In addition, the algorithm inherits the randomness and stable tendency of the cloud model, so it not only keeps the population diversity, but also avoids the precocity. It protects the optimal individual, so to a great extent it overcomes the defect of the poor local search capability, lack of memory and slow convergence in the traditional genetic algorithm.