Discovery-driven Exploration of OLAP Data Cubes
  Sunita Sarawagi      Rakesh Agrawal      Nimrod Megiddo
----------------------------------------------------------

A dominant application of OLAP data cubes is the interactive
exploration of business data by an analyst to find regions of anomaly.
Analysts often have to go through multiple drill-downs and selection
stages before finding anything interesting.  We propose a new
discovery-driven exploration paradigm that pre-mines the data for such
exceptions, summarizes the exceptions at appropriate levels, and leads
the analyst to the interesting regions.  We discuss the statistical
basis of our approach and describe computation techniques that make
this methodology feasible for large data bases.