Abstract: To solve the problem that the feature weights are difficult to quantify accurately, a hybrid algorithm based on grey wolf optimizer (GWO) algorithm and bird swarm algorithm (BSA) was proposed to optimize the feature weights. First, Chebyshev map, opposition-based learning and elitism strategy were used to initialize the population of the hybrid algorithm. Second, the location updating formula of GWO algorithm and the foraging behavior of BSA were combined as the improved location updating strategy of the algorithm for local search. Then, the vigilance behavior and flight behavior of BSA were integrated into the hybrid algorithm to obtain a balance strategy for global search. A convergent grey wolf and bird swarm algorithm (GWBSA) was obtained, and the feature weights were optimized through the iteration of GWBSA. Experiments were carried out by using benchmark functions and standard classification data sets, respectively. Compared with the genetic algorithm, the ant lion algorithm and other algorithms, the GWBSA has fast convergence speed and is hard to fall into local optimum, which can improve the solution quality of pattern classification problems.