As the world enters a new era of big developments and changes, the inter-country and inter-region scientific collaboration has seen fundamental changes in its environment, which not only faces unprecedented opportunities but also encounters grand challenges. Consequently, exploring the network dynamics and driving mechanisms of global scientific collaboration plays a critical role in the judgment of global science competition and cooperation. This paper, from the perspective of "node-line-network" and based on the dynamic logic in proximity approaches, uses the internationally co-authored papers data from the InCites database of Clarivate Analytics company for the period 2000-2019, and investigates the position of countries/regions, bilateral partnerships between countries/regions, the evolution of network structure, and reveals dynamic mechanisms of international scientific collaboration network as well as how the influence of driving factors changes over time. Firstly, results show that the traditional science powerhouses, such as the United States, the United Kingdom, France, and Germany, have occupied central and critical positions in the network. At the same time, the emerging scientific nations have strengthened their centrality and influence in the network, including China, South Korea, India, and South Africa, which are gradually close to the core position. Secondly, the number of ties between countries has increased dramatically. The research partnership between Germany and the United States is gradually giving way to China-US collaboration, which becomes the most important bilateral collaboration between countries in the world. Thirdly, the global scientific collaboration network is characterized by a star-shaped structure. As globalization and networking of science advance, decentralization in the collaboration network grows increasingly refined. The whole network has evolved from a single group dominated by the United States to a double group including the United States and Saudi Arabia. Finally, the extended gravity model indicates that geographical proximity, social proximity, cognitive proximity, common language, historical links, and talent flows have a positive effect on international scientific collaboration. As the knowledge network evolves, the importance of geographical proximity, common language, and historical links has waned over time, while the significance of social proximity, cognitive proximity, and talent flows has increased.