Doctor of Philosophy
Electrical and Computer Engineering
Dr. Luiz Fernando Capretz
Delay of Publication
In this research, we propose a novel model to predict software size and effort from use case diagrams. The main advantage of our model is that it can be used in the early stages of the software life cycle, and that can help project managers efficiently conduct cost estimation early, thus avoiding project overestimation and late delivery among other benefits. Software size, productivity, complexity and requirements stability are the inputs of the model. The model is composed of six independent sub-models which include non-linear regression, linear regression with a logarithmic transformation, Radial Basis Function Neural Network (RBFNN), Multilayer Perceptron Neural Network (MLP), General Regression Neural Network (GRNN) and a Treeboost model. Several experiments were conducted to train and test the model based on the size of the training and testing data points. The neural network models were evaluated against regression models as well as two other models that conduct software estimation from use case diagrams. Results show that our model outperforms other relevant models based on five evaluation criteria. While the performance of each of the six sub-models varies based on the size of the project dataset used for evaluation, it was concluded that the non-linear regression model outperforms the linear regression model. As well, the GRNN model exceeds other neural network models. Furthermore, experiments demonstrated that the Treeboost model can be efficiently used to predict software effort.
Bou Nassif, Ali, "Software Size and Effort Estimation from Use Case Diagrams Using Regression and Soft Computing Models" (2012). University of Western Ontario - Electronic Thesis and Dissertation Repository. Paper 547.
Available for download on Sunday, May 25, 2014