Systèmes d'Information et Management

Article Title

Prédiction d' effort pour les projets informatiques. Un modèle économétrique


The aim of this paper is to present software estimation models using a bottom-up approach. Our prediction models have the following characteristics : - they take into consideration several variables related to project logic (inputs, outputs, transaction types, etc.), project technical requirements and performance drivers ; - they use data collected at the early stages of the project (conceptual and logical models) as well as data describing the final stages of the project (physical model, programming, testing and installation). The prediction models have been validated on data related to several different projects of a large French financial institution. The projects have been developed in a technical environment using a relational database system, a fourth-generation language within a client-server computer architecture. The projects have been managed under the SDM/S project management and MERISE methodology. We show that the ex-post models, presented in this paper, lead to a realistic effort estimation. These seven models rely on linear regression based on only one or two independant variables. During the validation process, we have shown that the uncalibrated COCOMO model does not fit well this technical environment. The predicted values based on functional points analysis (FPA) are better than the results obtained by COCOMO. Our model and FPA give similar results.

This document is currently not available here.