Enhancing sparrow search algorithm via multi-strategies for continuous optimization problems

As a recent swarm intelligence optimization algorithm, sparrow search algorithm (SSA) is widely adopted in many real-world problems. However, the solutions to the limitations of SSA (such as low accuracy of convergence and tendency of trapping into local optimum) are still not available. To address these issues, we propose an enhanced multi-strategies sparrow search algorithm (EMSSA) based on three strategies specifically addressing the limitations of SSA: 1) in the uniformity-diversification orientation strategy, we propose an adaptive-tent chaos theory to allow more diversity and greater randomness in the initial population; 2) in the hazard-aware transfer strategy, we construct a weighted sine and cosine algorithm based on the growth function to avoid trapping into the state of local optima stagnation; 3) in the dynamic evolutionary strategy, we design the similar perturbation function and introduce the triangle similarity theory to improve the exploration capability. The performance of EMSSA in solving the continuous optimization problems about the 23 benchmark functions, CEC2014, and CEC2017 problems is much improved than that of SSA and other state-of-the-art algorithms. Furthermore, the results of the density peak clustering optimization show that the EMSSA outperforms SSA.