Database evaluation using multiple regression techniques

Abstract

A model of the inverted file of an automated bibliographic system is constructed using the Zipf distribution of word frequency. By ascertaining the parameters of the Zipfian model of the inverted file system, one can estimate the minimum data storage requirements of the database. In addition, given a few additional system parameters, access time for a specified query can be estimated. The estimation procedures are accomplished using logarithmic transformations and multiple regression techniques. This paper introduces the Zipfian models, their regression formulation, and their results and interpretation for application to database evaluation.