A Fitness-Based Evolving Network for Web-APIs Discovery

Abstract

Web-APIs such as Google-Maps, Twilio and Twitter APIs form the building blocks of many mobile and web-based applications. They enable cross-organizational functionality integration and data sharing over the Web. Tens of thousands of these Web-APIs with a wide spectrum of functionalities are currently available on various registries such as ProgrammableWeb.com. However, despite the continuous addition of new Web-APIs of various qualities to these registries, only a few, with certain preferences that do not necessarily define quality, are consistently discovered and used by service consumers. The main reasons for this meagre uptake have been identified as the isolation of web-APIs, poor scaling mechanism and lack of social connection among these APIs. Even though, existing Web-APIs discovery approaches show promising results, this task continues exacerbate service engineers. In this paper, we propose a fitness-based complex network approach for connecting Web-APIs into a global network to facilitate Web-APIs discovery. We construct the network based on the theoretical procedure of Bianconi-Barabási complex network model. The procedure is considered in two phases: First, we estimate Web-APIs' fitness using the Random Walk algorithm and define a strategy for capturing their popularity. Second, we grow the network based on each API's fitness and popularity. Using the popular ProgrammableWeb datasets, we evaluate the APIs network using the universal network metrics and compare our network topology with that of real networks. Finally, we show how our network can be used to facilitate Web-APIs discovery. The results present in this work are expected to serve as practical guide for modelling evolving-network-based Web service solutions, particularly, service discovery and recommendation applications.