In order to improve the performance of social spider optimization(SSO) algorithm, a social spider optimization algorithm with the dynamic learning strategy(DSSO) is proposed. In this algorithm, a dynamic selection mechanism for the learning factor in population cooperation is applied to balance solution accuracy and search speed. A manner to update individual combining randomized crossover strategy and cloud theory is proposed to improve the collaboration manner, which can maintain the diversity of the population as much as possible and improve searching speed. Experimental results on benchmark functions show that the DSSO algorithm improves convergence property and robustness compared with the representative four algorithms.