A Methodology to Analyze and Estimate the Software Development Process Using Machine Learning Techniques

Abstract

Analyzing the software development process and estimating the effort required for its completion is an essential task. In the case of Agile methodology, the values of the parameters used for estimation vary frequently as the scope of the project changes with changes in the requirements of the clients. Hence, the estimation done at the initial phase will not be appropriate until the completion of the project. Therefore, to overcome this issue, a methodology is proposed to estimate the duration of a project by applying machine learning techniques. The use-case point method is used for estimating the duration. Information about the number of use cases and values for environmental and technical factors is stored in a repository. Few values may be uncertain, and to estimate the effort for a new project with few unknown or uncertain values, the machine learning algorithm Gaussian Process Regression (GPR) is used. The repository information is taken as the training dataset, and the new project data is taken as the test dataset. The estimated value shows the accurate duration for the new project. The result is validated with a popular dataset.