In view of the disadvantages of slow convergence speed and weak local search ability of the artificial bee algorithm, this paper proposes a new improved artificial bee algorithm. The honey source is sorted according to the fitness value, and then the sorting result is used as the weight value to construct a virtual honey source, namely the weighted center. If the weighted center is superior to the current optimal solution, the current optimal solution is replaced to obtain a better current optimal solution. On the basis of the weighted center, the full-dimension search strategy is added to improve the local search capability of the algorithm. The application of two strategies speeds up the convergence speed of the algorithm and enhances the local search capability. The validity of the new algorithm on the classical 22 benchmark test function is analyzed. The experimental results show that the proposed algorithm can significantly improve the accuracy and speed. Under a given equivalent time, the solving accuracy and the convergence speed of the proposed algorithm are much higher than that the comparison algorithm.