Abstract: To solve the problems of the original sparrow search algorithm, such as poor global search ability, weak local development ability, and easy fall into local optimal, this paper proposes an adaptive t-distribution sparrow search algorithm (ATSSA) based on adaptive t-distribution. Tent chaos mapping is introduced to initialize the population to increase the diversity of the initial population. Subsequently, the adaptive t-distribution mutation operator is used to perturb individual positions to improve the global search ability. Meanwhile, the t-distribution mutation operator introduced by dynamic selection probability adjustment is used to balance the global search ability of the algorithm. Finally, the elite reverse learning strategy is integrated, and the location generated the optimal solution is disturbed to generate a new solution, which prompts the algorithm to jump out of the local optimal. Ten benchmark test functions are used in the simulation experiment, and the results show that ATSSA has better optimization ability than SSA. Moreover, ATSSA and deep extreme learning machines are used to construct a prediction model. The octane data set is used to carry out the experiments, and the prediction accuracy of the model is improved from 87.31% to 99.32%, verifying the algorithms promising potential for engineering applications.

Key words: sparrow search algorithm, Tent chaotic mapping, adaptive t distribution, dynamic selection strategy, elite reverse learning