Statistical analysis serves as a foundation for informed decision making in a number of areas central to IBM's business, such as quality assurance, manufacturing, pricing, marketing, business intelligence and delivery of services. Extracting useful information from the ever increasing quantities of data generated in today's business environments presents significant challenges. IBM Research has traditionally been a leader in several areas of statistics, in particular in forecasting, time series analysis, reliability and process control, linear modeling and risk analysis. Critical advances in business forecasting techniques, Web traffic prediction, and risk management methods have come from our labs. Shaping appropriate data collection and modeling paradigms to enable practical decision support tools are key goals of current work in the area of statistics. Some of our recent highlighted research projects include:
- Disaster Risk for Computer Systems (see image on right)
- Analytics for Software Development Decision Support
- Statistical models for the VLSI route problem
- Component Cost Forecasting and Product Planning
- Multiscale Prediction of Web Server Capacities
- Predictive Models for Strategic Outsourcing
- Hard Disk Drive Reliability Analysis
IBM researchers continue to be leaders in advancing the state-of-the-art in statistical methodology. Being part of IBM gives us unique opportunities to have our research affect both the methodological and practical components of statistical research. We advance the methodology by publishing in leading journals and conferences, producing patents, and collaborating with researchers from top-ranked academic institutions. We affect statistical practice through significant engagements, both with business units that make up IBM, and also with key IBM customers.
Related Publications
Christiansen, W.F. (Southern Methodist University), Amemiya and Y. (IBM Research). Latent variable analysis of multivariate spatial data. Journal of the American Statistical Association 97:302-317, 2002.
Guo and Xin (IBM Research). Some Risk Management Problems for Firms with Internal Competition and Debt. Journal of Applied Probability 39:55-69, 2002.
Hong, Se.June (IBM Research), Hosking, Jon (IBM Research), Natarajan and Ramesh (IBM Research). Multiplicative Adjustment of Class Probability: Educating Naive Bayes. Proceedings of INFORMS 2002. 2002.
Lee and Jon (IBM Research). Maximum entropy sampling. In Encyclopedia of Environmetrics, by A.H. El-Shaarawi and W.W. Piegorsch, editors, Wiley, 2001.
Ray, Bonnie (IBM Research), Tsay and R. (Univ. Chicago). Bayesian Methods for Change-Point Detection in Long-Range Dependent Processes. Journal of Time Series Analysis 23:687-705, 2002.
Yashchin, E. (IBM Research), Wisniewski, M. (IBM Research), Franch, R., Conrady, D., Fiorenza, G., Noyan and C.. Estimating the efficiency of collaborative problem-solving, with applications to chip design. IBM Journal of Research and Development 47:77-88, 2003.
Frequency of Flood Incidents in 1979-2002: Total = 473
This graph shows the relative risk, as a function of location, that a computer system will be put out of action by a disaster. Disasters include both natural hazards such as hurricanes, earthquakes and floods, and internal events such as CPU failure or A/C failure. The graph is part of a statistical risk analysis conducted for IBM Business Continuity and Recovery Services, which uses the results to ensure the availability of its disaster recovery services.