There are two popular approaches when developing a software system, the first being System Analysis/System Design (SA/SD) and the second being an Object-Oriented (00) approach. In either approach a poor design inevitably produces a poor application. Thus, being able to evaluate the quality of a systems design is advantageous.The Design Metrics Research Team at Ball State University has developed metrics that measure the quality of software systems. From this research the metric D(G) (a composite of the metrics De and D) has been shown to be very effective when used to determine fault prone modules in a system. However, D(G) has been primarily applied to systems developed using the SA/SD approach. This thesis translated D(G) to an 00 setting and empirically evaluated D(G) to determine if D(G) is a good predictor of error-prone classes. The results indicate that De and Di are indeed good predictors of error-prone classes. Of the classes highlighted by De, 67% had errors, while 100% of the classes highlighted by D; had errors. / Department of Computer Science
| Identifer | oai:union.ndltd.org:BSU/oai:cardinalscholar.bsu.edu:handle/185871 |
| Date | January 1997 |
| Creators | Cox, Jeffrey A. |
| Contributors | Zage, Wayne M. |
| Source Sets | Ball State University |
| Detected Language | English |
| Format | v, 53 leaves : ill. ; 28 cm. |
| Source | Virtual Press |
Page generated in 0.0018 seconds