An elite symbiotic organisms search(SOS) algorithm based on subpopulation stretching operation is proposed to solve the problems of premature and slow convergence in SOS. In the mutualism phase, the population is divided into two subpopulations according to the fitness value: One is responsible for development and the other is aimed at exploration. The targeted evolutionary strategy is designed for each subpopulation, which makes the algorithm keep a good balance between convergence speed and accuracy. In the commensalism phase, the individual updating mode is modifred by using the directional information of the optimal individual and introducing the stretching factor and the difference perturbation vector, which improves the convergence speed of the algorithm and guarantees the diversity of the population. The ``parasitism' mechanism of the elite, which is proposed by simulating the biological relationships between parasites and hosts, keeps a further balance between development and exploration. The comparison examinations of the standard SOS, the improved SOS and other four intelligent evolutionary algorithms on 17 functions indicate that the proposed algorithm has better accuracy and obvious advantage of convergence speed.