Determining the Topic Evolution and Sentiment Polarity for Albinism in a Chinese Online Health Community: Machine Learning and Social Network Analysis

BackgroundThere are more than 6000 rare diseases in existence today, with the number of patients with these conditions rapidly increasing. Most research to date has focused on the diagnosis, treatment, and development of orphan drugs, while few studies have examined the topics and emotions expressed by patients living with rare diseases on social media platforms, especially in online health communities (OHCs). ObjectiveThis study aimed to determine the topic categorizations and sentiment polarity for albinism in a Chinese OHC, Baidu Tieba, using multiple methods. The OHC was deeply mined using topic mining, social network analysis, and sentiment polarity analysis. Through these methods, we determined the current situation of community construction, identifying the ongoing needs and problems experienced by people with albinism in their daily lives. MethodsWe used the albinism community on the Baidu Tieba platform as the data source in this study. Term frequency–inverse document frequency, latent dirichlet allocation models, and naive Bayes were employed to mine the various topic categories. Social network analysis, which was completed using the Gephi tool, was employed to analyze the evolution of the albinism community. Sentiment polarity analysis was performed using a long short-term memory algorithm. ResultsWe identified 8 main topics discussed in the community: daily sharing, family, interpersonal communication, social life and security, medical care, occupation and education, beauty, and self-care. Among these topics, daily sharing represented the largest proportion of the discussions. From 2012 to 2019, the average degree and clustering coefficient of the albinism community continued to decline, while the network center transferred from core communities to core users. A total of 68.43% of the corpus was emotional, with 35.88% being positive and 32.55% negative. There were statistically significant differences in the distribution of sentiment polarity between topics (P