Progressive OLAP (POLAP)
Mediator architecture that integrates time-series data from several web sites and databases from agencies such as the Bureau of Labor Statistics (BLS), the Census Bureau, and the Energy Information Administration (EIA).We describe how we extend this architecture by incorporating a local data warehouse to allow for more efficient processing of OLAP queries.
Multidimensional databases are a powerful tool for data warehousing and mining. Starting with range-sums, we are developing new techniques for the progressive evaluation of aggregate queries on multidimensional databases. Progressive techniques provide quick approximate query results that become more accurate as query evaluation progresses. By giving users useful information long before the exact result is available, progressive algorithms can provide interactive response times where exact algorithms cannot. We achieve this with our algorithm POLAP (Progressive OLAP) by using the wavelet transformation to decompose a query into a sum of sub-queries. This technique yields an exact range-sum algorithm that has the best known query/update complexity. By evaluating the most important sub-queries first, we obtain excellent approximate results after a small fraction of the total I/O required by the exact algorithm. POLAP extends well to statistical databases with continuous dimensions. Using POLAP in conjunction with wavelet-based density estimators of a data density function, we obtain statistically meaningful aggregate query results with error estimates.