Software Effort Estimation Models Incorporating Expert Judgment

Project management is important to avoid software development failure. Effort estimation is the basis of the management. A mathematical model estimates development effort based on software size and productivity. Productivity factors are can be used as explanatory variables only when they are measured qualitatively or quantitatively. We propose that a project manager judge productivity level roughly (high or low), and use it as an explanatory variable, instead of using unmeasurable productivity factors as explanatory variables. However, the productivity level includes some errors. For instance, an estimation target project may be judged as high productivity level erroneously, although actual productivity level will be low. So, we propose alternative method considering misjudgment. In an experiment, we changed error of productivity level and compared proposed methods with a conventional model. As a result, our method is robust for the error of productivity level, and the method is more accurate than the conventional model in most cases.