Multi-Label Symptom Analysis and Modeling of TCM Diagnosis of Hypertension
Traditional Chinese Medicine (TCM) has been used for diagnosis of hypertension and has significant advantages. Symptom analysis and modeling of TCM provides a way for the clinician to produce a service to users to accurately and efficiently diagnose hypertension. In this study, an ensemble learning framework based on network clustering analysis with information fusion is proposed. We first analyze the frequency distribution and cluster heat map of TCM hypertension clinical cases, and establish a network based on the syndrome and symptom of cases. Through the analysis of community networks, we get the dominant and subordinate syndrome and construct a sub-classifier to co-train and improve the performance of the classifier. Then we use ML-KNN and RAkEL-SVM multi-label classifiers to train and test the cases. Considering the result of 10-fold cross validation, we discover that ML-KNN and RAkEL-SVM with information fusion have better performance than traditional learning methods without information fusion. For all evaluation criteria, the average precision of ML-KNN is higher, and the F-Measure does not vary substantially. But the averaged recall of RAkEL-SVM is significantly higher.
2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)
(2019). Multi-Label Symptom Analysis and Modeling of TCM Diagnosis of Hypertension. 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM).
Available at: https://aquila.usm.edu/fac_pubs/15840