A Pluggable Architecture for Building User Models From Massive Datasets

In many situations, it is common that a large single source of data serves as input to multiple application areas, each of which may use a different user model. It is often the case that each user model is assembled using a different process, however, in general, it is more efficient to have a single architecture for building different user models for different application areas. We propose an architecture based on MapReduce that allows for processing terabytes of information in a timely fashion using pluggable components, each one capable of including different features in the final user model. A metamodel is used for specifying the characteristics of the desired user model, which can include a short-term user model and a long-term user model, and the architecture is responsible for building it from the specified data and pluggable components. We present an instantiation of the architecture for telecommunication (telco) applications and evaluate how the architecture escalates with a real dataset. Our evaluation indicates that complex user models for millions of users can be obtained in a timely fashion.