Project Longview
Longview: Querying the Future Now
Summary
Our goal is to develop data management 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 or missing 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 on large-scale data sets or high-rate data streams 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 project 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. We also expect that the resulting technology will also allow for a deeper understanding and support for user-defined functions in database systems.
Participants:
Students:
Faculty:
Publications:
- On Predictive Modeling for Optimizing Transaction Execution in Parallel OLTP Systems.
Mert Akdere, Ugur Cetintemel, Matteo Riondato, Eli Upfal, Stanley Zdonik. In ICDE 2012 (to appear). - Modeling and Prediction of Concurrent Query Performance.
J. Duggan, U. Cetintemel, O. Papaemmanouil, E. Upfal. In SIGMOD 2011 (to appear). - The VC-Dimension of SQL Queries and Selectivity Estimation through Sampling.
M. Riondato, M. Akdere, U. Cetintemel, E. Upfal, S. Zdonik. In ECML-PKDD 2011. - The Case for Predictive Database Systems: Opportunities and Challenges.
M. Akdere, U. Cetintemel, M. Riondato, E. Upfal, S. Zdonik. In CIDR’11. - Database-support for Continuous Prediction Queries over Streaming Data.
M. Akdere, U. Cetintemel, E. Upfal. In PVLDB Volume 3(1), 2010.
Acknowledgements:
The Longview project is supported by the NSF grant IIS-0905553.
Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.