Developing Software Size Estimating Relationships Based on Functional Descriptions of the Software.
Abstract
This thesis researched the ability to develop regression models to predict the number of source lines of code (LOC) based on functional descriptions of the software. LOC, a major cost driver in currently available software cost estimating models, has been consistently underestimated, thus lowering not only the software cost estimate but also the total cost estimate of the weapon system. Six software sizing data bases containing various functional variables were used. The variable included complexity, reliability, experience level of programmers, etc. For each data base, regression analysis was performed to derive the optimal model to predict LOC. Of the five data base containing complexity, it was statistically significant in three. The best developed model was for Armament Division's airborne computer programs. The correlation coefficient sq R was .6583 for the two variables in model. These were; (1) the system for which the program was developed and (2) the reliability needed in the program. The initial research has been accomplished, but more data and further research is needed.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1986
- Accession Number
- ADA174335
Entities
People
- Mark J. Whetstone
Organizations
- Air Force Institute of Technology