.png){kind=icon}
.png){kind=icon}
Return
Generative adversarial networks-based software development effort estimation for small datasets
.png){kind=icon}
OA
AI
DOI：10.1016/j.sciaf.2025.e03156.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In project management, accurately estimating the effort required for software projects is crucial. With the right configuration, artificial neural networks (ANNs) can be used to build predictive models since they learn from previous data and minimize errors. In this study, we combined the Feed-Forward Backpropagation Neural Network algorithm (FBNN) with a Generative Adversarial Network (GAN) to estimate the effort of software development projects and address the challenge of small datasets by generating new samples. Besides, we investigated the impact of using three scaling methods (min-max scaler, robust scaler, and standard scaler) on the performance of the GAN-FBNN model using six datasets (COCOMO81, NASA93, Desharnais, Kitchenham, ISBSG, and Maxwell). We evaluated and compared the performance of the GAN-FBNN model applying the aforementioned scaling methods with that of a simple FBNN model. Statistical tests were also used to compare and rank the four models. The experimental results demonstrate that combining GAN and data scaling with the FBNN model in SDEE is promising, showing better performance than the classical FBNN model.
Keywords:
Feedforward backpropagation neural network
Generative adversarial network
Software effort estimation
ANN
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
3.3
Papers:
1.0K
Citations:
6.7K