Abstract: As the simple genetic algorithm has the shortcomings such as low convergence velocity and efficiency, and easily falling into premature convergence, a kind of improved adaptive crossover operator and mutation operator was introduced.Together with a simple genetic neural networks model (SGA－BP) being built, an improved adaptive genetic neural networks model (IAGA－BP) was built using a decimal encoding scheme.By comparison of the analysis results between the two models, it showed that the SGA－BP model was more excellent in the convergence velocity and the precision.