Knowledge discovery meets linked APIs

Abstract

Knowledge Discovery and Data Mining (KDD) is a very wellestablished research field with useful techniques that explore patterns and regularities in large relational, structured and unstructured datasets. Theoretical and practical development in this field have led to useful and scalable solutions for the tasks of pattern mining, clustering, graph mining, and predictions. In this paper, we demonstrate that these approaches represent great potential to solve a series of problems and make further optimizations in the setting of Web APIs, which have been significantly increasing recently. In particular, approaches integrating Web APIs and Linked Data, also referred to as Linked APIs, provide novel opportunities for the application of synergy approaches with KDD methods. We give insights on several aspects that can be covered through such synergy approach, then focus, specifically, on the problem of API usage mining via statistical relational learning.We propose a Hidden Relational Model, which explores the usage of Web APIs to enable analysis and prediction. The benefit of such model lies on its ability to capture the relational structure of API requests. This approach might help not only to gain insights about the usage of the APIs, but most importantly to make active predictions on which APIs to link together for creating useful mashups, or facilitating API composition.