To improve the indoor visible light positioning accuracy, we proposed a visible light fingerprint positioning method using the least squares support vector machine (LSSVM) optimized by the beetle antenna search (BAS) algorithm. First, the signal strength characteristics of the LED light intensity were used to build the fingerprint database. Next, the BAS algorithm was applied to optimize the hyperparameters of the least square support vector machine to improve accuracy and reduce time cost. Finally, we obtained the mapping relationship between the position coordinates and signal strength characteristics to achieve positioning. The experimental results show that the BAS-LSSVM positioning method can achieve a good positioning effect, with the positioning error of 97.0%, test points being less than 0.10 m, and the average positioning error of all test points being 0.031 m.