Support Vector Regression을 이용한 소프트웨어 개발비 예측

The purpose of this paper is to propose a new software development cost estimation method using SVR (Support Vector Regression). SVR, one of machine learning techniques, has been attracting much attention for its theoretic clearness and good performance over other machine learning techniques. This paper may be the first study in which SVR is applied to the field of software cost estimation. To derive the new method, we analyze historical cost data including both well-known overseas and domestic software projects, and define cost drivers affecting software cost. Then, the SVR model is trained using the historical data and its estimation accuracy is compared with that of the linear regression model. Experimental results show that the SVR model produces more accurate prediction than the linear regression model.