MCDB-strongRstrong Risk Analysis in the Database.pdfVIP

MCDB-R: Risk Analysis in the Database Subi Arumugam Ravi Jampani Luis L. Perez Fei Xu Christopher Jermaine Peter J. Haas University of Florida Rice University Microsoft Corporation IBM Research - Almaden Gainesville, FL, USA Houston, TX, USA Redmond, WA, USA San Jose, CA, USA sa2,jampani@cise.ufl.edu lp6,cmj4@ feixu@ phaas@ ABSTRACT large investment loss. The problem is made more complex by the Enterprises often need to assess and manage the risk arising from fact that the tails of the distribution are hard to specify a priori, and uncertainty in their data. Such uncertainty is typically modeled as so risk analysis frequently focuses on the inverse problem of deter- a probability distribution over the uncertain data values, specified mining an extreme quantile, e.g., determining the value such that by means of a complex (often predictive) stochastic model. The there is a 0.1% probability of seeing a loss of or more. Such a probability distribution over data values leads to a probability dis- “value at risk” can be viewed as a probabilistic worst-case scenario, tribution over database query results, and risk assessment amounts and might be used to specify precisely where the “upper tail” of the to exploration of the upper or lower tail of a q