Un modelo paramétrico matemático difuso para la estimación del esfuerzo de desarrollo del software

Abstract

Effort estimating techniques of product development constitute an imperative for any industry area. Financial staff require estimation activities in the context of feasibility analysis of projects. Mathematical parametric estimation models are nowadays some of the most frequently used estimation technique, since they provide a sound empirical basis to predictions. Nonetheless, existing software estimation parametric models lack an explicit support for information imperfections like imprecision and uncertainty, which are inherent to the linguistic formulation of cost divers. Existing research on the introduction of imperfection handling techniques in such models does not provide comprehensive model formulations, and areas like cost diver aggregation are neglected. The research effort described in this document aims at providing an extended parametric software estimation model that provides support for imperfection in the assessment of the inputs and in the adjustment of the model itself, applying for that purpose concepts related to Fuzzy Set Theory. Concretely, a crisp parametric model is extended to allow for the use of fuzzy numbers as inputs, and a concrete fuzzy regression technique is studied as an approach for the adjustment of the model. In addition the use of membership elicitation techniques as an approach to model inputs is also studied, and an analysis in provided regarding the adjustment of aggregation operators as a model for the contribution of second-level aspects to the input.