A syndrome differentiation model of TCM based on multi-label deep forest using biomedical text mining

Area of Science:

• Computational intelligence
• Biomedical informatics
• Traditional Chinese Medicine (TCM)

Background:

• Syndrome differentiation is fundamental to TCM diagnosis and treatment.
• Modernizing TCM requires scientific, intelligent diagnostic methods.
• Challenges in TCM data include standardization issues and patient privacy, hindering large-scale dataset creation.

Purpose of the Study:

• To develop an intelligent method for TCM syndrome differentiation.
• To address the limitations of small and non-standardized TCM datasets.
• To improve the accuracy and efficiency of TCM diagnostic models.

Main Methods:

• Proposed a multi-label deep forest model (ML-PRDF) incorporating an improved multi-label ReliefF feature selection algorithm.
• Enhanced feature representativeness and reduced feature dimensionality.
• Optimized the model for high classification accuracy with reduced data processing costs, particularly for small sample sizes.

Main Results:

• The ML-PRDF model demonstrated superior performance over other multi-label classification models.
• Achieved higher accuracy in TCM syndrome differentiation compared to traditional multi-label deep forest models.
• The PCC-MLRF feature selection algorithm proved effective in selecting representative features.

Conclusions:

• The proposed ML-PRDF model offers an effective solution for intelligent TCM syndrome differentiation, especially with limited data.
• The model enhances diagnostic accuracy and efficiency in TCM.
• Feature selection plays a crucial role in optimizing deep forest models for TCM applications.