Abstract: Aiming at the randomness of pelican optimization algorithm in solving problems, a feature selection method based on multi-strategy fusion pelican algorithm is proposed. Firstly, we use the best point set theory to initialize the population instead of the random strategy in the original pelican algorithm, so that the population distribution is uniform and the ergodicity is improved; Secondly, the reverse differential evolution algorithm is used to optimize the population individuals after each update iteration to improve the global search performance; Finally, the adaptive t-distribution mutation strategy is used to perturb the optimal solution to prevent it from falling into the local optimum, and six standard test functions are selected for simulation. Experiments show that the improved algorithm can select the optimal features more effectively and improve the classification accuracy than other algorithms.