API Learning: Applying Machine Learning to Manage the Rise of API Economy

Abstract

Application Programming Interface (API) exposes data and functions of a software application to third-party users. In digital business, API economy is one of the key component for determining the value of provided services. With the rise in number of publicly available APIs, understanding each API endpoint manually is not only labor intensive but it is also an error prone task for software engineers. Due to the complexity of understanding the sheer number of APIs, it is difficult for software developers to find the best possible API combinations (i.e. API Mashups). In this demonstration, we introduce API Learning platform which employs machine-learning based technologies to efficiently search APIs, validate APIs, and generate API mashups. These technologies enable a machine to automatically generate machine-readable API specification from API documentations, understand variety of APIs, validate extracted information through automated API validation, and finally recommend API mashups for a specific purpose. As of now, API Learning platform collected over 14,000 API documentations and generates a machine readable format for REST APIs with an accuracy of 84%. The proposed demo prototype shows how it enables users to quickly find relevant APIs, automatically verify API availability, and get the best possible API mashup recommendations.