Estimating object-relational database understandability using structural metrics

Abstract

New Object-Relational Database Management Systems (ORDBMSs) are replacing existing relational ones. In spite of the high expressiveness, application systems built upon ORDBMS are more complex and difficult to maintain due to the mixing of two paradigms, the relational and the objectoriented. This paper describes a suite of metrics for measuring different aspects of an object-relational database. An empirical validation of the usefulness of the proposed metrics in estimating the understandability of an object-relational schema is given. The analysis procedure comprises the use of two techniques: C4.5, a machine learning algorithm, and RoC, a robust Bayesian classifier. The results demonstrate that a subset of the proposed measures is relevant for the estimation of the understandability.