Approximate query answering in numerical databases

This work addresses the problem of efficient processing of queries in very large numerical databases. Previous focus has been on the design of index structures for the efficient access of data. Recently more and more statistical methods have been used in query optimization. Those methods approximate the distribution of the attribute values to estimate the selectivity of query results. A methodology that uses regression techniques to approximate the actual attribute values is introduced. Through analysis of the data, one derives a set of characteristic functions to form a "regression database", a compressed image of the original database. Based on these functions, approximate answers to queries may be provided within a pre-specified tolerable error, but without the expensive search overhead usually inherent with the use of indexing techniques. A framework to build regression databases is proposed. An experimental prototype is implemented to evaluate the technique in terms of realizability, efficiency and practicality. This technique is complementary to conventional approaches and to statistical methods.