Towards Traditional Chinese Medicine Symptom Prediction with Multi-label Learning Method

Artificial intelligence technology has driven further exploration and application of the value of traditional Chinese medicine (TCM). However, TCM diagnostic methods face challenges in practical application, such as the coexistence of multiple symptoms in patients, which requires diagnostic methods to consider the complexity of these symptoms comprehensively. Multi-label classification algorithms can effectively identify and classify TCM symptoms. Therefore, we constructed a TCM symptom prediction framework based on multi-label learning method. To be specific, we collect TCM data through inspection, smelling, inquiry, and palpation. Next, we perform data preprocessing through feature scaling, label coding, and data partitioning. Then, we learn the specific data representation for each label by incorporating second-order label correlations. Finally, experiments on the two TCM datasets show that the proposed method significantly outperforms state-of-the-art methods and provides a new perspective for TCM research.