STATISTICAL COST-MODELLING OF FINANCIAL TIME SERIES FUNCTIONS

摘要

We present a statistical regression approach to building a cost model of an aggregate financial time series function. The cost model is needed by an object-relational DBMS query optimizer. This approach is much easier than the traditional analytical approach and yet achieves a highly precise model. Users need only provide a set of variables influencing the costs. This requires only high-level understanding of how the function works. Experiments show that the cost models thus built are highly precise and that quadratic models are adequate.