Leveraging Social Support Types and Link Prediction for User Recommendation for Online Health Community in Pregnancy After Loss

Abstract

For those pregnant after a prior pregnancy loss, receiving support can be crucial in coping during this emotionally tumultuous time. One of the ways to receive support is through online health communities. Users reach out to online health communities for informational and nurturant support. Since responses are crowdsourced, there is no way to influence the content of messages. However, we want to ensure that users receive the best support suited for their individual needs. Therefore, this research revolves around a prediction algorithm that utilizes a combination of the network structure of the group and the social support content of the messages to determine the best user to respond to another user. Using data from the Pregnancy After Loss subreddit, we utilize our model to predict whether two users will interact. Through this prediction algorithm, find that the combination of these two algorithms is complimentary and results in better performance. We also better understand the content of messages in the group and how they can be further classified to different subtypes of social support.