The evolution of keywords co-occurrence network based on MAG dataset, 1800–2017

The Microsoft Academic Graph (MAG), a sizeable heterogeneous graph of academic entities containing 166,192,182 scientific documents reported by 201,493,854 authors from 17,885 affiliations over two hundred years from 1800 to 2017, is utilized in this paper. Here, a big data-driven analysis method is developed to study the keywords of scientific articles. We find 94,476,176 articles containing 113,608 different keywords. Furthermore, based on the keyword database, we construct annual co-occurrence keywords network (weighted network) and analyze the statistical and dynamic evolution patterns, such as strength distribution, average distance, clustering coefficients, and rich-club phenomena. From the knowledge of complex network, we try to reveal the research hotspots of discipline, which can help us to understand this new product of Microsoft Academic and its safe use.