JMIR Publications

ABSTRACT

Background:

With the rapid development of the application of internet, various types of online medical health communities continue to appear on the internet, providing a platform for internet users to share experiences and ask questions about medical and health issues. The convenience of internet services and trust in online health information may affect user search behaviour. However, there are problems in the online medical health community, such as the low accuracy of the classification of user question data and the uneven health literacy of users, which affect the accuracy of user retrieval and the professionalism of professional medical personnel answering.

Objective:

The paper proposes a multilevel classification framework based on Graph Convolutional Network(GCN) model for user needs in online medical and health communities. Most online medical health communities tend to provide only disease type labels, which do not provide a comprehensive summary of user needs. The study aims to solve this challenges which various unclassified or misclassified information interferes with the targeted answers of medical professionals and the information retrieval results of other users.

Methods:

The paper takes the Chinese online medical health community of "Qiuyi" as an example and crawls the user question data of the "Cardiovascular Disease" section as the data source. First, the disease types involved in the problem data are segmented by manual recognition method to generate the first-level label, and then the needs are identified by K-means clustering method to generate the user information needs label as the second level label. By constructing a graph convolutional network model, user question data are automatically classified, thus realizing the multilevel classification of user needs.

Results:

The paper provides empirical research about user information needs are automatically classified in Chinese online medical health community. Based on the evaluation metrics of accuracy, precision, recall, and F1-score, the proposed classification model is proved to have better performance.

Conclusions:

A multilevel classification framework has be designed based on Graph Convolutional Network(GCN) model. Through the method, disease type classification and information demand classification are carried out so that users can conduct more targeted information retrieval.