ABSTRACT A PARAMETERIZED MODEL FOR SELECTING THE OPTIMUM FILE ORGANIZATION IN MULTI-ATTRIBUTE RETRIEVAL SYSTEMS

This thesis develops a general parameterized model that facilitates the comparison of different file organization techniques for a given multiple key information retrieval system. The model is based on minimizing the expected pro-cessing time of the data base in performing on-line retrieval and updating operations. The decision rules are a function of the relevant characteristics of the data base, the on-line queries, the storage devices, and the file organization tech-niques, as well as the relative breakdown of the processing requests between retrievals and various types of updating operations. To demonstrate the use of the model, detailed timing formulas are developed for the retrieval and updating opera-tions for three different file organizations: the Multilist system, the Inverted Index system, and the Cellular Seria