Enabling the definition and reuse of multi-domain workflow-based data analysis

Abstract

Data analysis applications have become essential to extract significant insight from heterogeneous data sources. However, their development requires technical expertise in computer science techniques like data mining, making its broad adoption by non-experts difficult. In this context, workflows have emerged as a high-level solution to define and automate the sequence of steps involved in the data analysis process, hiding the low-level computational requirements. Existing workflow systems have some difficulties related to their complexity to adapt the provided elements and their inability to reuse workflow definitions. To address these problems, a novel framework for creating customized, ready-to-use and interoperable workflow systems is proposed and prototyped in this paper. Its multi-layer architecture has been designed on the basis of the separation of concerns and the reuse of knowledge assets. As a result, the presented approach allows reducing the time-to-market and saving development costs.