Node occlusion and edge congestion problems, which are caused by the increasing scale and complexity of network, had become a hot spot in network visualization research. To solve the visual clutter problem in network, edges close to each other in network were bundled by curving them. A segmental forced-directed algorithm (FDA) simplification model and a community based compatible edge bundling network model were proposed and improved. To solve the problem of excessive bending of some edges in segmental FDA bundling model, network was divided into different communities by CNM cluster algorithm. Experimental result shows that the bundling simplification algorithm introduced above has a wide applicability, and network visualized by this algorithm has good visual effect and readability.