According to nonlinear secondorder twopoint boundary problem, a hybrid genetic algorithm based on real coding is constructed by the combination of the genetic algorithm with LevenbergMarquardt algorithm. Because the former has the advantages of global optimization ability while the latter has strong local optimization ability, therefore, the improved algorithm not only has global optimization ability but also the calculation accuracy can not be affected by initial values, can use less computing time, and can effectively improve the convergence speed of the algorithm. Finally, the comparison between analytical solution and exact solution to nonlinear secondorder twopoint boundary value problems by using the improved algorithm indicates that the improved algorithm has big advantages of nonlinear secondorder twopoint boundary value problems computation and is an effective method for numerical solution.