User-Constrained Clustering In Online Requirements Forums

[Context & motivation:] Software development projects involving geographically dispersed stakeholders often use web-based discussion forums to gather feature requests. Our previous study showed that users have a tendency to create redundant threads as well as large unfocused mega-threads. [Question/problem:] In this paper we propose novel solution for integrating user feedback into the process of dynamically and iteratively clustering features into discussion threads. [Principal ideas/results:] We integrate feed back in the form of sticktogether and move-apart advice, plus user-defined tags into our consensus based clustering process. [Contribution:] Experimental results demonstrate that our approach is able to deliver high quality and stable clusters to facilitate forum-based requirements elicitation.