The project consisted in the development and implementation of a system capable of ranking feeders according to their likelihood of failure in real-time. The system used a series of machine learning techniques ranging from boosting to online learning, and required a significant amount of software engineering. For more information on the project, check out the following publications and this link:

• Real-time Ranking with Concept Drift Using Expert Advice. H. Becker, M. Arias. KDD 2007, Aug 2007.
• Predicting Electricity Distribution Feeder Failures Using Machine Learning Susceptibility Analysis. P. Gross, A. Boulanger, M. Arias, D. L. Waltz, P. M. Long, C. Lawson, R. Anderson, M. Koenig, M. Mastrocinque, W. Fairechio, J. A. Johnson, S. Lee, F. Doherty, A. Kressner. AAAI 2006, Jul 2006.

This project was partly supported by a research contract from Consolidated Edison Company of New York.

The project consisted in the developtment of an integrated system to process in real time a huge incoming stream of alerts produced by current intrusion detection systems. A key component of this system includes an unsupervised clustering algorithm that combines a temporal sliding window, entropy tests, and expert rules to track the on-the-fly evolution of alert groups. More information can be found at this link.

This project was partly supported by a PROFIT grant from the Ministerio de Ciencia e Innovación of Spain.