Software time estimation using regression methods

In actual scenery of the development industries, making good estimates is essential for the survival of organizations. Estimating much more than necessary can lead loss of new contracts and doing on the contrary can cause huge financial losses. Generating efficient time estimates is one of the fundamental points in software projects because it helps the clients to see how much time will be spent to develop the project through a schedule. Thus, this paper presents the performance of regression methods for software projects time estimation: linear regression, parametric quantile regression, and nonparametric kernel regression. The performance of the methods is assessed by the mean magnitude of relative errors (MMRE). Experiments were carried out using twelve projects data set from NASA repository. The results showed that kernel regression provides a versatile method of exploring a general relationship between variables and gives good predictions of software programming time yet to be made without reference to a fixed parametric model.