In this paper, we present a new approach to partitioning communities in a complex network via degree of dependence of nodes. We define the dependence degree of a node on its neighbors, the dependencetce degree and the conditional dependence degree of a node on a cluster. The main point of the approach is to partition the nodes, which have the biggest dependence degree and are only dependent on nodes, firstly to clusters, then to absorb nodes whose dependence degree or conditional dependence degree on cluster gets the right value, until all the nodes are partitioned to the right communities. The partition of our approach in some real-world network satisfies the definition of communities, and in the network whose communities are already known, our partition method fits the physical truth.