Exploring Health-Related Topics in Online Health Community Using Cluster Analysis

Recently patients are increasingly turning to online health community to share their experiences and exchange healthcare knowledge. Exploring hot topics from online health community helps us better understand their needs and interests in health-related knowledge. However, statistical-based topic analysis employed in previous studies is becoming impractical to process the growing large-scale online data. Automatic topic analysis based on document clustering is an alternative approach but usually produce poor results as a result of lack of domain-specific knowledge. So this paper proposes a novel framework for health-related topic analysis using text clustering integrating medical domain-specific knowledge. Experiment results show that adding medical domain-specific features into feature set could achieve significantly better clustering performance than existing methods. In addition, further analysis reveals that there also exist some significant differences about hot topics among different kinds of disease discussion boards.