A Classification Approach to Detect Social Support in Online Health Communities.

Online Health Communities (OHCs) is a term that is commonly used to refer to online social networks related to health. There are two types of users in an OHC, namely authors and visitors. OHC authors are primarily patients and caregivers who use the platform to share their health journeys, learn about illnesses, seek social support, and connect with others in similar circumstances. OHC visitors are family members and friends who are following the authors’ journeys and offering social support. Different types of social support may be sought and provisioned in an OHC including prayer, esteem, emotional, network, instrumental, and informational. In this project, we employed classification algorithms to automatically identify the type of social support provisioned by an OHC visitor where this identification of social support can be used by a platform to enrich its services by, for example, identifying patterns that trigger more support provisioning, and by identifying which type of support keeps patients and caregivers more engaged on the platform. Identifying the social support provisioned in an OHC post is a multi-label classification problem since a single post may involve more than one social support type. Hence, we implemented a multi-label classifier as a set of independent binary classifiers, one for each of the following five social support types: esteem, emotional, network, instrumental, and informational. Each binary classifier is used to detect the existence or the absence of the corresponding support type in a given OHC post. Four of our classifiers are at least 85% accurate while the informational support classifier’s accuracy is 77%.