The quality of evaluation index has strong coupling correlation and data dependency in the evaluation of network community. To solve the problem of poor quality of traditional single evaluation compared with optimized network community discovery algorithm, this paper proposes online community discovery algorithm based on multi-objective particle swarm optimization. The algorithm generates Pareto optimal community classified collection through the optimization of multiple online community quality evaluation indicators at the same time, in which users can choose the most satisfied community structure according to the specific needs. Finally the comparison experiment is carried out between the single objective optimization method and multi-objective optimization algorithm. The experimental results show that the proposed algorithm can dig out higher quality online community in the absence of priori information and have higher stability of the system.