The BP neural network is improved by PSO algorithm had solved the disadvantages of BP neural networks algorithm itself, such as slow convergence, non-convergence. But the PSO algorithm has it self’s problems, such as precocity and local convergence, so the EO algorithm is introduced, and the BP neural network is improved by the combination of EO algorithm and PSO algorithm. The practice result indicates that the BP neural network improved by PSO-EO has good convergence and higher forecast accuracy, better than the traditional BP algorithm and BP algorithm optimized by PSO algoritm.