A multi-surrogates algorithm is developed to deal with the mixed-integer programming problems. Firstly, a sampling method based on the model of the multi-swarm PSO is developed to ensure the accuracy and diversity of the the samples. Furthermore, the local surrogate models are constructed by an online modeling method based on the data parallel approach. Then, the collaborative optimization is carried out based on the preselecting strategy and PSO. Finally, the effectiveness of the proposed method are verified by the 14 test problems and 1 data driven model parameter selection problems.