| Summary: | 碩士 === 國立交通大學 === 資訊管理研究所 === 83 === Software testing is crucial to assure the quality of software
systems. Cost estimation of testing is therefore very important
for feasibility study and scheduling of system development.
This research intends to find a convenient and improved model
for estimating the cost of software testing. The research first
collects 17 real projects of software testing in the last five
years. Based on past theories, possible variables affecting the
cost of testing are decided. Then the technique of factor
analysis is used to filter out important variables that are
independent of each other. Finally, nine factors stand up after
regression and collinearity analysis. These nine factors are:
complexity, reliability, interface, defect rate, modern
programming skill, tool, schedule demand, test cycle, and test
experience. This result shows that domestic testing is very
different from that of the United States. Comparing our model
with COCOMO model, our model is noticeably simpler and more
accurate.
|
|---|
