PORTNEUF - A Framework for Continuous User Involvement
Published on 21. October 2013
Book
Paperback/Softback
231 pages
978-3-8439-1218-1 (ISBN)
Description
Post-deployment user feedback such as feature requests and bug reports have the potential to improve software quality and reveal missing functionality. Nevertheless, in practice software companies often ignore user feedback and simply stick with predefined roadmaps and development plans, because of two main problems. First, in order to benefit from user feedback, developers have to analyze, consolidate, and structure it, which becomes time-consuming when feedback volume is high. Second, in order to prioritize their work, developers need to assess how many users are affected by a particular feedback, by identifying duplicate and similar feedback in a manual way.
This thesis describes Portneuf, a framework which tackles both problems. It consolidates user feedback by employing a context-aware, hybrid recommender system. Moreover, it introduces FeedbackRank, an algorithm which allows developers to assess the importance of collected user feedback to the user community.
We demonstrate the feasibility and applicability of our approach in two real-world applications. In a quasi-experiment with a large number of users, we showed that Portneuf increases developers' efficiency when working with user feedback. The framework significantly reduces the amount of user feedback duplicates by over 67%, with the recommender system showing a hit-rate of over 82%. Moreover, we obtained an average precision of around 76% for FeedbackRank, which suggests that it provides a valuable estimation of what is important to the user community.
This thesis describes Portneuf, a framework which tackles both problems. It consolidates user feedback by employing a context-aware, hybrid recommender system. Moreover, it introduces FeedbackRank, an algorithm which allows developers to assess the importance of collected user feedback to the user community.
We demonstrate the feasibility and applicability of our approach in two real-world applications. In a quasi-experiment with a large number of users, we showed that Portneuf increases developers' efficiency when working with user feedback. The framework significantly reduces the amount of user feedback duplicates by over 67%, with the recommender system showing a hit-rate of over 82%. Moreover, we obtained an average precision of around 76% for FeedbackRank, which suggests that it provides a valuable estimation of what is important to the user community.