In order to improve location precision of three-dimensional wireless sensor nodes, a novel three dimensional node location method of wireless sensor network is proposed in this paper based on least squares support vector machine (LSSVM) which parameters are optimized by artificial fish algorithm (AFSA). Firstly, the study samples are constructed for three-dimensional nodes localization model, and then LSSVM is used to build three-dimensional node localization model in which fish feeding behavior, cluster and rear end behavior are simulated to find the optimal parameters of LSSVM, and finally the performance is tested by simulation experiment. The results show that, compared with other localization methods, the proposed method can improve the precision of the sensor nodes and it has some practical application values.