多层次特征融合的中医药材推荐方法研究

Research on the Recommendation Method of TCM Herbs Based on Multi-level Feature Fusion

深度学习技术迅速发展,在中医药材推荐任务中被大量使用。针对传统神经网络模型在中医药材推荐应用中推荐精度不高、模型参数量较大等问题,提出一种多层次特征融合的轻量级药材推荐方法。在TextCNN模型参数量少、特征抽取全面等特点的基础上,进一步融合症状语义特征和序列特征,从而获取更全面的症状药材特征完成中医药材推荐任务,并将其在中医药材公开数据集上进行验证。实验表明,该方法对药材推荐的F5得分达到0.241 9,相较于基线模型具有显著提升,模型大小仅为4.26M。并且,通过消融实验分析不同模型组件对推荐任务的影响,验证了所提方法的有效性,以期为中医药材推荐提供新的方法。

Deep learning technology of online attack and defense experimental courses has developed rapidly, and has been widely used in the task of recommending traditional Chinese medicine. Aiming at the problems of low recommendation accuracy and large model parameters of traditional neural network model in the application of traditional Chinese medicine recommendation, a lightweight drug recommendation method based on multi-level feature fusion is proposed. On the basis of the features of TextCNN model, such as less parameters and comprehensive feature extraction, the symptom semantic features and sequence features are further integrated, so as to obtain more comprehensive characteristics of symptomatic medicinal materials to complete the task of recommending traditional Chinese medicine, and it is verified on the public data set of traditional Chinese medicine. The experiment shows that the F5 score recommended by this method for medicinal materials reaches 0.2419, which is significantly improved compared with the baseline model, and the model size is only 4.26M. In addition, the ablation experiment was conducted to analyze the impact of different model components on the recommendation task, which verified the effectiveness of the proposed method, with a view to providing a new method for the recommendation of traditional Chinese medicine.