The National Science Foundation awarded a research grant in the anticipated amount of $1.2M to Ugur Cetintemel, Eli Upfaland Stan Zdonik to develop database technology that would simplify building predictive analytics applications over large-scale data. Predictive analytics involves analyzing historical and current data to make predictions about future data values, events, and trends, and has a wide range of applications in security, marketing, economics, sociology, genetics and computing. The generic predictive database technology to be developed will make computing with predictions easier to express and far more efficient than the prevalent application-level solutions that are known to be brittle and unscalable.
The concrete product of the work will be a new type of database system, called Longview, that seamlessly integrates predictive models as first-class primitives by intelligently incorporating them in the process of data management and query optimization. Longview will develop novel algorithms, data structures and interfaces to automatically load, train, select, and execute predictive models. The project will also investigate "white-box" model support, in which the knowledge of the semantics and representation of models, if available, will be used to enhance the quality and performance of predictions. The team expects that the resulting technology will also allow for a deeper understanding and support for user-defined functions in database systems.