The Study On Genetic Algorithm Based On Space Search

Genetic algorithm is one of the earliest evolutiona ry algorithm, it has well stability and global search capability, widely used in practical problems. Although compared with the current particle swarm and differential evolution algorithm, it has shortcomings in local optimization and convergence speed. We commitment to the theoretical basis for research of genetic algorithm, build different genetic algorithm model, analysis its convergence and provides a good foundation. We will genetic algorithm combined with a variety of differe nt mechanisms or put forward a new improvement strategy, increase the application domain algorithm, improve the efficiency of algorithm.In this paper, based on spatial search way, we could have a knowledge of the species distribution in the solution space, and put forward the improvement strategy of genetic algorithm for research analysis.The main content of this paper is as follows:1) Study the theoretical basis of genetic algorithm, and analyzes its convergence process. Genetic algorithm is a kind of parallelism algorithm ba sed on heuristic search. It has a good optimization ability and simple process. From the schema theorem, we can learn that during the genetic algorithm, it is hard to retain longer model. They have a bigger chance of being damaged. Crossover and mutation operation is to make individual coding can be random distribution in the variable space. In the process of guided by fitness value, only the some coding can keep relatively stable individual. But the genetic algorithm would trap in premature. This paper mainly puts forward a way to produce new individuals and keep the diversity of population.2) In single objective genetic algorithm, based on the adaptive genetic algorithm, we propose a space partitioning strategy. In order to avoid the adaptive ge netic algorithm trap into local optimum in the late and improve the efficiency of searching, This paper proposes a space division by the weight of the individuals in the population distribution range, so as to accelerate the convergence process. In the process of iterative genetic algorithm, it need to statistical analysis of distribution of population, know population distribution state of range and observe the process of convergence. Improved adaptive genetic algorithm understand the distribution of the population in the whole variable space state, redistribute in the part of the population, increase the diversity of individuals to accelerate the convergence process. We could found that, the improved adaptive genetic algorithm has the difference on the diversity of population, at the same time, can quickly converge to global optimal solution through the experiment.3) In the multi-objective genetic algorithm, we propose the strategy of space decision tree. In high dimension space, we could not choose the better space solution set. Then, we record each individual value and keep the individual in the evolution of the relative stability of partial weight spanning tree construction. Though generating populations of spatial decision tree to guide the search direction, we can effectively guarantee the individual in the process of optimization to keep a certain distance with diversity, and quickly accelerate to the global search. We could find that in the NSGA2 algorithm, using the space decision tree optimization problem can be obtained better effect for the target dimension higher high-dimensional multi-objective through the experiment.