Application of mathematical quantification theory class III in software effort prediction

Prediction of software development effort, commonly relies on historical data. In recent years, the software development environment become increasingly diverse, and a wide variety of variables are recorded in project data. In addition to quantitative variables such as function points, project data tends to record many qualitative variables, such as development language and platform. However, there has not been any significant discussion on how to treat this large quantity of qualitative variables. In this paper, an effort prediction method using mathematical quantification theory class III is proposed to facilitate the handling of qualitative variables. From the results of an experimental comparison with conventional methods, we learn that the predictive accuracy can be improved by applying the mathematical quantification theory class III to effort prediction.