A Statistical Cost-Modeling of Financial Time Series Functions for an Object-Relational DBMS Query Optimizer

Financial time series functions are in prevalent use in stock market analysis, and are so important in business applications as to be supported by every major commercial object-relational database management system (ORDBMS). These ORDBMSs require users to provide the cost functions of user-defined functions (UDFs) for their query optimizers. The traditional approach to developing a cost function is to build an analytic function of variables derived from data configurations and system configurations. This is an overwhelming task for users without advanced knowledge of the internal implementation of an ORDBMS. In the case of financial time series functions, there is a statistical approach much easier for users. Users provide a set of variables influencing the costs (as we call the cost variables) based ∗The contact author, e-mail: [email protected], tel: (802)656-1919, fax: (802)656-0696