The Experimental Analysis of the Efficiency of Genetic Algorithm Based on 3-satisfiability Problem

Genetic algorithm is widely used to solve the satisfiability problem in recent years. However, it still suffers from premature convergence and loss of precise. With bubble sort combining dichotomy selection, we studied 3-SAT problem on genetic algorithm. We used C++ programming language to simulate the algorithm. In this paper, we evaluated the computational efficiency of the genetic operators and related parameters on the genetic algorithm. A larger number of experiments were conducted. The results showed that the success probability S with crossover is larger than that of without crossover. Crossover plays an important role on S. The performance of the genetic algorithm arises as the variables and population size increases. When the population size is larger, the effect of variables on success probability is stronger. Meanwhile, there are some connection between the success probability S and the the generation of the first hitting G. When the mutation rate P-m <= 0.1, the curve of P-m-dependence of 1/G is similar to the P-m-dependence of S.