Open-Sourcing Software: An Effort Estimation Model Open Access
Downloadable ContentDownload PDF
Large or mid-size Internet companies are unable to efficiently assess whether to open-source a closed source project or to continue to invest in their existing open-source project, which results in higher operating costs.This research proposes a model called iOSSEstimator, which efficiently predicts effort and forecasts the effort savings percentage (ESP) for a proposed or existing open-source project hosted on Github.com. Statistical and machine learning methods were used to extend and improve on existing open-source effort prediction models. Sixty-four popular open-source projects consisting of 11,497 versions and 6.4 million tasks were used to support development of the iOSSEstimator model. The results of this research indicate that existing open-source effort estimation models can be augmented to improve effort estimation with a mean accuracy PRED25 range of 82% to 97% across nine different maintenance task types, such as bug fixing, refactoring, and testing. This enables the calculation of the effort savings percentage which helps inform the decision to open source projects or to continue to invest in an existing open-source project. From the 64 examined projects only 58% had a positive ESP. The high percentage of open-source software projects with negative ESP further supports the need for employing this tool. The tool can be used to help guide a decision-maker to open-source a closed-source project or to continue investing in an existing open-source project.