Abstract: To solve the problems of low diversity, slow convergence speed and easy to fall into local extremum of the lion swarm optimization algorithm, an improved adaptive algorithm based on multi-strategy is proposed. Specifically, the adaptive parameters are introduced to improve Tent chaotic map for population initialization, which ensures random distribution and improves diversity and uniform ergodicity. Then, based on differential evolution mechanism, the disturbance factor of lioness position update is introduced to enhance the ability of the algorithm to jump out of the local optimum. Finally, the second order norm and information entropy are combined to form a step size disturbance factor, which adaptively adjusts the selection probability of different behavior modes of the cub, therefore inhibits the premature convergence of the algorithm. Based on adaptive Tent chaotic search, individuals with poor fitness are improved through multiple neighborhood points of local optimal solution to further enhance the optimization speed and accuracy. Comparing with various intelligence algorithms, the better performance of the proposed algorithm is verified through 16 multi-type test functions. To further evaluate the effectiveness of the proposed algorithm, it is used to optimize the initial weights and thresholds of back propagation neural networks. Experimental results on the two datasets show the proposed algorithm has higher classification accuracy compared with the other three algorithm.

Key words: lion swarm optimization algorithm, multi-strategy, tent chaotic map, differential evolution, information entropy, adaptive Tent chaotic search