Social question answering (SQA) services allow users to clarify their queries by asking questions and obtaining answers from other users. To enhance the responsiveness of such services, one can identify similar questions and, thereafter, return the answers available. However, identifying similar questions is difficult because of the complex language structure of user-generated questions. For this reason, we developed an approach to cluster similar questions based on a web of social relationships among the questions, the answers, the askers, and the answerers. To do so, we designed a graph-based cluster analysis using design science research guidelines. In evaluating the results, we found that the proposed graph-based cluster analysis is more promising than baseline methods.