Towards Adaptive Analytics on Big Data Sources

The analysis of Big Data is a core and critical task in multifarious domains of science and industry. Such analysis needs to be performed on a range of data stores, both traditional and modern, on data sources that are heterogeneous in their schemas and formats, and on a diversity of query engines. The users that need to perform such data analysis may have several roles, like, business analysts, engineers, end-users etc. Therefore a system for Big Data analytics should enable the expression of analytics tasks in an abstract manner, adaptable to the user role, interest and expertise. We propose a novel workflow model that enables such users to define in an abstract manner the application logic of the analysis of diverse Big Data. The model focuses on the separation of task dependencies from task functionality. Our motivation and applications derive from real use cases of the telecommunication domain.