Abstract: In this paper, we provide an adaptive gravitational search algorithm based on DBSCAN. During iterations the initial population is clustered into certain number of groups, and then individuals in subgroups update velocity, position and mass values by their own fitness and gravitational constant values. Individuals adaptively update their information. Clustering and adaption are adopted to improve global search ability and convergence rate. What's more, it provides effective balance between exploration and exploitation. Experimental results show that better performance has been achieved with high convergence rate and better convergence accuracy by using the improved GSA.