afast -AT- cs.umass.edu

In my dissertation research, I am developing state-of-the-art automated techniques for correctly and efficiently identifying causal mechanisms underlying complex social, biological, and technological networks. Large, interconnected, dynamic networks such as the social network among fraudulent securities dealers in the United States or peer-to-peer music file sharing networks present many challenges that hinder the identification of causal mechanisms. First, the complexity and structure present in these network data can lead to learning incorrect model structure. My work includes an exploration of techniques for improving the correctness of learned models through a utilization of effect size, statistical power and computationally intensive statistics. Second, due the size of these data more efficient algorithms for causal identification are imperative. I am also exploring methods from fields outside of computer science such as genetic epidemiology to prune the space of possible causal models leading to more efficient identification of cause and effect relationships. Applications of my work include automated detection of securities fraud in the US securities industry, modeling users in peer-to-peer music file sharing networks, citation analysis of academic fields, and predicting the playoff performance of NFL head coaches.

My contributions will be released as part of the Proximity open-source software package developed by the Knowledge Discovery Laboratory. Proximity is developed using extreme programming methodology including test-driven development and pair programming and is available at http://kdl.cs.umass.edu/proximity.

Journal Articles

Exploiting relational structure to understand publication patterns in high-energy physics.  McGovern, A., L. Friedland, M. Hay, B. Gallagher, A. Fast, J. Neville and D. Jensen.  SIGKDD Explorations 5, 2 (December 2003), 165–172. Winning Entry: KDD Cup 2003 [PDF] [Abstract] [bibtex]

Conference and Symposium Articles (Peer Reviewed)

Relational Data Pre-Processing Techniques for Improved Securities Fraud Detection. Fast, A., L. Friedland, M. Maier, B. Taylor, D. Jensen, H. Goldberg, and J. Komoroske. In Proceedings of the 13th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. (2007). 941-949. [PDF] [Abstract] [bibtex] [Video of the Presentation]

The NFL Coaching Network: Analysis of the Social Network Among Professional Football Coaches. Fast, A., and  D. Jensen. In AAAI Fall Symposium on Capturing and Using Patterns for Evidence Detection. (October 2006) [PDF] [Abstract] [bibtex] [2006 Playoff Predictions]

Creating Social Networks to Improve Peer-to-Peer networks. Fast, A. , D. Jensen and B. N. Levine. In Proceedings of the 11th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. (2005). 568–573. [PDF] [Abstract] [bibtex]

Workshop Articles, Technical Reports, and Other Publications

Understanding the effects of search constraints on structure learning. Hay, M., A. Fast and D. Jensen. University of Massachusetts Amherst, Technical Report 07-21. (2007). [PDF] [Abstract] [bibtex]

Learning Models of Macrobehavior in Complex Adaptive Systems. Fast, A. AAAI/SIGART Doctoral Consortium 2006. [PDF] [bibtex]