Browsing by Author "Ndenga, Malanga Kennedy"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Assessing quality of open source software based on community metrics
    (International Journal of Software Engineering and Its Applications, 2015) Ndenga, Malanga Kennedy; Méhat, Jean; Ivaylo, Ganchev; Wabwoba, Franklin
    The purpose of this study is to analyze data from Open Source Software (OSS) community with an objective of identifying community metrics that can predict quality of OSS projects. We experimented with data from Apache OfBiz and Apache httpd-2 server OSS projects. We applied linear regression technique to the dataset to assess the strength of possible relationships of variables and also examined possible trends amongst variables. From the analysis, we found out that the size of user mailing list has a correlation with number of reported bugs. We concluded that the size of user mailing list community may not be an accurate representation of the entire user community that adopted the project basing on quality. However Backlog Management Index was found to be a better metric for assessing how projects manage issues reported by users.
  • Thumbnail Image
    Item
    Performance and cost-effectiveness of change burst metrics in predicting software faults
    (Knowledge and Information Systems, 2018) Ndenga, Malanga Kennedy; Ganchev, Ivaylo; Mehat, Jean; Wabwoba, Franklin; Akdag, Herman
    The purpose of this study is to determine a type of software metric at file level exhibiting the best prediction performance. Studies have shown that software process metrics are better predictors of software faults than software product metrics. However, there is need for a specific software process metric which can guarantee the best fault prediction performances consistently across different experimental contexts. We collected software metrics data from Open Source Software projects. We used logistic regression and linear regression algorithms to predict bug status and number of bugs corresponding to a file, respectively. The prediction performance of these models was evaluated against numerical and graphical prediction model performance measures. We found that change burst metrics exhibit the best numerical performance measures and have the highest fault detection probability and least cost of misclassification of software components.