Software development productivity prediction of individual projects applying a neural network

Machine learning techniques have been applied in the software engineering field and their models could be applied for predicting the development productivity of software developers. In this paper, a neural network model was trained from a data set of 140 individual projects developed from between years 2005 and 2008 with practices based on a process specificaly designed to laboratory learning environments: Personal Software Process (PSP). Then, this model was applied for predicting the productivity of a new projects consisting of 156 projects developed from between years 2009 and 2010. The code in all projects was developed by 74 graduated students, using object oriented programming languages C++ and Java. Prediction accuracy obtained from neural network was compared to those obtained from a fuzzy logic model as well as from a statistical regression model. Results suggest that a neural network model could be used for predicting development productivity of individual projects, when they are developed in a disciplined way in a laboratory learning environment