Comparison of Typical Clustering Algorithms for Patent Citation Network

The comparative study of clustering algorithms for patent citation network can provide effective help for the practical application of citation network clustering. This method extracts and constructs citation network from patent document data, and the stable period of network development is identified by scale-free effect. On this basis, the optimal clustering algorithm is selected from a variety of typical clustering algorithms by using modularity index, Calinski-Harabaz index and community coverage. An empirical study was conducted using 9121 patent records in flexible display field from 1963 to 2018 in the Derwent Innovations Index. The results show that Leiden clustering based on modular optimization has the best effect in the application of patent citation clustering.