Abstract:With the layout results intuitive and easy to analyze, the network layout algorithm plays a critical role in network visualization based on the Force-Directed model. However, a high-quality layout result is not obtained easily by current network layout algorithms in a brief period when confronted with large-scale network data. An algorithm based on PageRank"s Force-Directed model is proposed in this paper, which can produce a better layout with aesthetic metrics such as Crosslessness , Minimum angle metric and so on. Moreover, to enhance the layout quality, the algorithm introduces PageRank to perfect the gravity and repulsion force calculation of nodes. Simultaneously, this paper proposes an adaptive step length based on PageRank to balance the efficiency and quality of the layout. Finally, a flexible CPU+GPU heterogeneous parallel computing framework was designed based on CUDA to effectively reduce the calculation time of the layout algorithm in the face of large-scale network data. The algorithm can produce a high quality layout via experiments with different types and sizes of network datasets. And under the same hardware conditions, the optimization scheme proposed in this paper is up to 58 times faster than the original algorithm.