MODELOS DESCRIPTIVOS Y PREDICTIVOS PARA LA ESTIMACIÓN DE COSTES EN PROYECTOS INFORMÁTICOS

Abstract

Due to their high uncertainty, the precise estimation of real costs in software projects is a very difficult task to achieve. The use of data mining and artificial intelligence techniques could be of great help to develop predictive and descriptive models in order to assist in this task. In this paper, several models obtained from the ISBSG Benchmarking & Research Suite Release 10 are presented. This database, provided by the International Software Benchmarking Standards Group, contains international software projects. The models were developed using the J48 y M5P machine learning algorithms that made possible to condensate the database knowledge into decision and regression trees. The database contains multiple project parameters: developing times and inactivity periods, projects quality, architecture types, costs, etc. From the models obtained, decision and regression trees easy to implement were developed. The results can help in the estimation of the software projects real costs according to their topology. Keywords: software projects; ISBSG; estimation of costs; estimation of time; data mining.