Evolving mashup interfaces using a distributed machine learning and model transformation methodology

Abstract

Nowadays users access information services at any time and in any place. Providing an intelligent user interface which adapts dynamically to the users’ requirements is essential in information systems. Conventionally, systems are constructed at the design time according to an initial structure and requirements. The effect of the passage of time and changes in users, applications and environment is that the systems cannot always satisfy the user’s requirements. In this paper a methodology is proposed to allow mashup user interfaces to be intelligent and evolve over time by using computational techniques like machine learning over huge amounts of heterogeneous data, known as big data, and modeldriven engineering techniques as model transformations. The aim is to generate new ways of adapting the interface to the user’s needs, using information about user’s interaction and the environment.