Detecting health misinformation in online health communities: Incorporating behavioral features into machine learning based approaches

Highlights：

• Based on the Elaboration Likelihood Model (ELM), the features of online health misinformation can be classified into two levels: central-level and peripheral-level.

• Four types of misinformation appear in online health communities: advertising, propaganda, misleading information, and unrelated information.

• We built a health misinformation detection model integrating the linguistic features, the topic features, the sentiment features, and the behavioral features.

• The proposed model, as well as the features, were validated on a real-world dataset, being able to correctly detect about 85% of health misinformation.

• The behavioral features are more informative than linguistic features in detecting health misinformation in online health communities.