Neural network and statistical modeling of software development effort

Abstract

Many modeling studies that aimed at providing an accurate relationship between the software project effort (or cost) and the involved cost drivers have been conducted for effectivemanagement of software projects. However, the derived models are only applicable for a specific project and its variables. In this chapter, we present the use of back-propagation neural network (NN) to model the software development (SD) effort of 18 SD NASA projects based on six cost drivers. The performance of the NN model was also compared with a multi-regression model and other models available in the literature.