Identifying Leaders in an Online Cancer Survivor Community

Abstract

Abstract Online communities are an important source of social support for cancer survivors and their informal caregivers. This research attempts to identify leaders in a popular online forum for cancer survivors and caregivers using classification techniques. We extracted user features from many different perspectives, including user-contribution, network, and semantic features. Based on these features, we further utilized the structure of the network among users and generated new neighborhood-based and cluster-based features. Classification results revealed that these features are discriminative for leader identification. Using these features, we developed a hybrid approach based on an ensemble classifier that performs better than many traditional metrics. This research has important implications for understanding and managing similar online communities. 1.