In order to overcome the problems of genetic algorithm, such as the low efficiency, genetic algorithm is redesigned by the complex system theory in this paper. First, the selecting operator is rebuilt by the power law, which is considered to be the self-organized criticality of complex system and sound distribution system of energy. Second, the crossover operator is redesigned by the characteristic of a self-learning complex system. Third, the generation strategy is improved by the mechanism of feedback. Finally, the gene floating operator is added to the algorithm. Because all operators are balanced with each other and restrict each other, the newly designed algorithm, complex system genetic algorithm (CSGA), improves efficiency and premature markedly. At last, experiments show that the CSGA is capable of dealing with high dimensional global optimization problems.