University of Southern California
    
Home   Contact Us    
Center for Systems and Software Engineering

About us
News
History
People
Events
Upcoming
Highlights
Past
Publication
Tech. Report
TR by Author
Research
Projects
Tools
Courses
Education
Degrees
Admissions
Affiliates
List of Affiliates
Private Area
Other Resources

Back        

Software Sizing

 

Personnel

TBD

 

Objectives

Research how the size of the software UML design impacts the software life-cycle effort. Build the COCOMO UML Extension, which will estimate the software project life-cycle costs from the software UML design.

 

Impacts

Traditionally COCOMO uses the number of equivalent source line of code (SLOC) as the major size driver for software life-cycle cost estimation. However, SLOC can only be accurately counted when the software construction is complete, while the most critical software estimations need to be performed before construction.

As an emerging industrial standard for object-oriented software analysis and design, UML has been widely used in presenting and visualizing software architecture. We believe UML metrics have a great potential to be a promising sizing method for next generation of software estimation tools because they can be used earlier in the project phase and are easier to build tools for automatic counting.

Our new research shows some promising potential for the estimators to use the UML sizing metric (e.g. the number of use cases) to estimate the software effort, duration and cost.

 

Approach

We counted the sizing statistics from the USC CSCI577 Software Engineering class projects in the past 3 years and analyzed the correlations among different sizing measures of the software project. As an initial result, Iwe found there is a significant statistical correlation between the number of use case scenarios and the number of source line of code in the projects we studied. Our conclusions are specific to the sample of 14 small eServices projects for which we have data on number of requirements, number of UML artifacts, and numbers of SLOC in various languages. However, they serve at least as a caution in calibrating expectations of the accuracy of UML-based sizing metrics at this stage in UML's definition and usage.

 

Publications

Yue Chen , Barry Boehm, Ray Madachy, Ricardo Valerdi, An Empirical Study of eServices Product UML Sizing Metrics , Accepted by the ACM-IEEE International Symposium on Empirical Software Engineering, August, 2004

Ricardo Valerdi, Yue Chen , Ye Yang, System Level Metrics for Software Development Estimation , Fast Abstract, Accepted by the ACM-IEEE International Symposium on Empirical Software Engineering, August, 2004

UML Sizing- 18 th International Forum on COCOMO and Software Cost Modeling

 

Future Plans

Perform controlled experiments, enhance the cost estimation tool, draft basic framework principles

 

If you have any questions about Software Sizing, please contact A. Winsor Brown at awbrown@csse.usc.edu.