A k-core Analysis to Large-Scale Web API Collaboration Networks

Abstract

The Web has become a huge programmable platform in the Web 2.0 era. More and more companies are opening their data and services to the public through Web APIs. With the increasing number and variety of Web APIs, new Web applications and value-added services can be rapidly developed by combining different APIs. The ever-growing collaborations between APIs thus raise a kind of a large-scale networks, namely Web API collaboration networks. However, the structure and evolution process of Web API collaboration networks are still unclear to people so far. This paper provides a deep analysis to the internal structure of a real-world Web API collaboration network using k-core decomposition. We firstly construct the Web API collaboration network by using the data crawled from the largest Web API registry, Programmable Web.com, and then employ the k-core decomposition method to obtain different subgraphs of the network with different centrality or coreness (i.e., k-cores). We give an experimental analysis to the structures of the Web API collaboration network and its k-cores by using some classic statistical tools, such as degree distribution and clustering coefficient. The analysis results not only can identify the most central APIs in the Web API collaboration network, but also provides a basis for the visualization and understanding of Web API collaboration networks.