Identification, Prediction, and Classification of Traditional Chinese Medicine Syndromes Based on Machine Learning Algorithms

The CNN-SVM AdaBoost fusion model constructed in this article combines the deep feature extraction ability of convolutional neural networks for multi-source four diagnostic information with the high-dimensional discriminative boundary advantage of support vector machines, and further corrects difficult cases using AdaBoost's iterative weighting mechanism to form an end-to-end syndrome identification system. Compared with classical logistic regression, gradient boosting, K-NN, and the best performing random forest, this model improved the accuracy from 87.4% to 95.8%, F1 score from 87% to 95.5%, AUC from 0.924 to 0.983 in the four classification task, with leading margins of 8.4%, 8.5%, and 0.059, respectively; Compared to the remaining baseline, there is a gap of more than 10%. What is particularly valuable is that the model has a high recall rate of 96% and an F1 score of 95.5%, which enables it to maintain sensitive capture and robust discrimination of minority categories in traditional Chinese medicine scenarios with fuzzy syndrome boundaries and imbalanced samples, significantly reducing the risk of misdiagnosis and missed diagnosis. Practice has proven that the collaboration between deep learning and traditional machine learning in the field of traditional Chinese medicine can break through the bottleneck of a single algorithm, provide reliable tools for accurate diagnosis and treatment, and have direct and profound significance for improving the quality of clinical diagnosis and treatment.