The basic chicken swarm optimization algorithm (CSO) is of slow convergence and easy to fall into local optimum. Aiming at the problems, we propose an improved constrained chicken swarm optimization algorithm (ICCSO) to improve the boundary constraint processing mechanism of the basic chicken swarm optimization, and the convergence speed and global search ability of the algorithm are also improved. Standard test functions and BP neural networks are taken as examples to demonstrate the rationality and effectiveness of the proposed algorithm.