Abstract: BP neural network is a hot research field in various disciplines for its powerful simulation calculation ability in recent years, but the algorithm has some shortages such as low convergence which limited the usage of the algorithm. Based on the genetic algorithm, the study improves the BP neural network to speed up its calculation. First, structure and shortages of BP neural network are analyzed for subsequent improvement; Second, the BP neural network algorithm is improved through parameter setting, initialization and evaluation, selection operator design, crossover and mutation operator design and algorithm calculation process design. Third, the improved algorithm is realized and the experimental results show that the improved BP algorithm has better performance in iterations and expectation fitting and can be used in internet public opinion evaluation practically.