Software cost and duration estimation based on distributed project data: A general framework

Abstract

Effort estimation is one of the most challenging tasks in the process of software project management. Enhancing the accuracy of effort estimation remains a serious problem for software professionals. Accurate estimation is difficult to achieve. The main difficulty is to collect distributed knowledge as data and information are often dispersed over different services, departments or organisations. Other main difficulty is to propose a model representative enough of this multi-partner behaviour. The objective of this study is to propose a general framework of the estimation starting from the analysis of the available projects database, the choice and establishment of estimation model, up to the use of this model to make estimation for new projects. In this paper, a comparative study between regression models and neural network models is performed. The proposed study is applied on a dataset of an automotive company.