A Multi-model Fusion Model for Traditional Chinese Medicine Syndrome Classification of Omicron Patients

Objective: As the new coronavirus evolves, the pathogenicity and impact of the Omicron strain on human health remain of concern. Traditional Chinese Medicine (TCM) has played a significant role in responding to this epidemic. This paper aims to examine the impact of TCM syndromes on clinical patient classification and provide references for personalized treatment planning by proposing a multi-model fusion (MMF) model for TCM syndrome classification. Methods: This study retrospectively analyzed the clinical characteristics of 900 confirmed Omicron strain infected patients in 9 cities in Shandong Province in the first half of 2022. The tongue and pulse conditions at the initial stage of infection are analyzed using the Spearman correlation coefficient method and mutual information method to select the initial features. The model is trained by integrating multiple learning strategies using the MMF approach. The optimal model is used to determine the essential components for classification. Results: The MMF model demonstrated better performance than Random Forest, eXtreme Gradient Boosting, and Support Vector Machine algorithms. The results of the feature importance analysis showed that TCM symptom features such as fever, cough, dry throat, hyperhidrosis, and tongue coating characteristics (greasy white, thick white, and white coating) are more prominent in distinguishing asymptomatic patients from mild patients. Conclusion: The MMF algorithm can improve the comprehensive research on TCM symptoms and the clinical classification of recent epidemics. The feature importance analysis provides insight into the indicators that should be given attention and serves as a reference for formulating a proper diagnosis and treatment plan.