Software Development Effort Estimation Using Regression Models Delve with Explainable Machine Learning (XAI)

Software projects often fail due to poor effort estimation, which can lead to issues such as wasted time and resources. Developing software requires significant time, money, and skilled people, making accurate effort estimation essential. Recent research uses machine learning algorithms and datasets to improve estimation, but these models can be difficult to interpret. In this study, we utilize SHAP (an explainable machine learning technique) to examine the key variables that impact effort estimation. We apply four different regression models to the China, Maxwell, and Desharnais datasets: Linear Regression, Random Forest Regression, Support Vector Regression and Artificial Neural Network. Our results identify the key attributes affecting effort estimation in each dataset, providing valuable insights into the models and improving their transparency and reliability.