| Resumen |
The Gamma Classi?er is a novel algorithm, immersed in the Associative Approach to Pattern Recognition, of which the Alpha-Beta BAM is another relevant model. The Gamma Classi?er has shown competitive performance in areas such as prediction of atmospheric pollutants, wireless network sensor location, and concrete mix properties forecast. This paper introduces the ?st successful application of this model to development effort prediction of software projects. In this sense, an ongoing concern of software managers is to predict how many hours should be spent on a development project, mainly regarding project budgeting and planning. Software managers based typically their predictions on judgment-based techniques; however, models-based techniques (statistical regressions, fuzzy logic, neural networks, or genetic programming) offer a good alternative. In this study, the Gamma Classi?er was trained with a data set of 163 software projects and then used for predicting the effort of another data set integrated by 68 projects; all projects were developed by 53 and 21 practitioners respectively. Accuracy result of this classi?er was compared with that of a fuzzy logic model and that from a statistical regression model.
|
