本論文針對根據適應症預測中醫藥方的挑戰，提出了一種結合知識圖譜與深
度學習技術的解決方法。透過建構知識圖譜來表示中草藥與適應症之間的複雜
關係，本研究提出了兩步驟的模型來增強藥方生成。主模型根據輸入的適應症
預測初始的中草藥組合，而輔助模型則考慮中草藥之間的互動，進一步完善這
些預測。這兩個模型相結合，旨在生成更具臨床相關性的藥方建議。
實驗結果顯示，該方法相比於現有模型在精確率、召回率以及排名品質方面
有顯著提升。特別是，輔助模型通過改進主模型的藥材組合，確保了藥材的有
效性及其協同兼容性。然而，輔助模型的表現仍受限於主模型的初始預測，這
限制了它在獨立學習複雜藥材互動方面的能力。
本研究通過使用知識圖譜驗證藥材推薦的相關性，為中醫藥方生成提供了一
種系統化的方法，進而提升了中醫治療的精準度和有效性，促進了中醫與現代
醫學的結合。

This thesis addresses the challenge of predicting Traditional Chinese Medicine
(TCM) prescriptions based on indications using a combination of knowledge graph
and deep learning techniques. By constructing a knowledge graph to represent
the complex relationships between herbs and indications, this research proposes a
two-step model to enhance prescription generation. The main model predicts an
initial set of herbs based on the input indications, while the auxiliary model refines
these predictions by considering herb interactions. Together, these models aim to
generate more clinically relevant herb recommendations.
Experimental results demonstrate the effectiveness of this approach, with sig-
nificant improvements in precision, recall, and ranking quality compared to existing
models. Notably, the auxiliary model improves the herb combinations suggested
by the main model, ensuring both herb efficacy and their synergistic compati-
bility. However, the auxiliary model’s performance is still influenced by the main
model’s initial predictions, limiting its ability to independently learn complex herb
interactions.
This research contributes to the integration of TCM with modern medical prac-
tices by utilizing a knowledge graph to validate herb recommendations, providing
a systematic approach to ensure the relevance of prescriptions. This approach
ultimately enhances the precision and efficacy of TCM treatments.

[1] Ehsaneddin Asgari and Mohammad RK Mofrad. Continuous distributed rep-
resentation of biological sequences for deep proteomics and genomics. PloS
one, 10(11):e0141287, 2015.
[2] Andrew L Beam, Benjamin Kompa, Allen Schmaltz, Inbar Fried, Griffin We-
ber, Nathan Palmer, Xu Shi, Tianxi Cai, and Isaac S Kohane. Clinical con-
cept embeddings learned from massive sources of multimodal medical data. In
Pacific Symposium on Biocomputing 2020, pages 295–306. World Scientific,
2019.
[3] Daniel Billsus, Michael J Pazzani, et al. Learning collaborative information
filters. In Icml, volume 98, pages 46–54, 1998.
[4] Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and
Oksana Yakhnenko. Translating embeddings for modeling multi-relational
data. Advances in neural information processing systems, 26, 2013.
[5] Nadav Brandes, Dan Ofer, Yam Peleg, Nadav Rappoport, and Michal Linial.
Proteinbert: a universal deep-learning model of protein sequence and function.
Bioinformatics, 38(8):2102–2110, 2022.
[6] Ahmed Elnaggar, Michael Heinzinger, Christian Dallago, Ghalia Rihawi,
Yu Wang, Llion Jones, Tom Gibbs, Tamas Feher, Christoph Angerer, Martin
Steinegger, et al. Prottrans: Towards cracking the language of life’s code
through self-supervised deep learning and high performance computing. arxiv
2020. arXiv preprint arXiv:2007.06225, 2007.
[7] Tymor Hamamsy, Meet Barot, James T Morton, Martin Steinegger, Richard
Bonneau, and Kyunghyun Cho. Learning sequence, structure, and function
representations of proteins with language models. bioRxiv, 2023.
[8] Yuanyuan Jin, Wendi Ji, Yao Shi, Xiaoling Wang, and Xiaochun Yang. Meta-
path guided graph attention network for explainable herb recommendation.
Health Information Science and Systems, 11(1):5, 2023.
[9] Yuanyuan Jin, Wendi Ji, Wei Zhang, Xiangnan He, Xinyu Wang, and Xiaoling
Wang. A kg-enhanced multi-graph neural network for attentive herb recom-
mendation. IEEE/ACM transactions on computational biology and bioinfor-
matics, 19(5):2560–2571, 2021.
[10] Thomas N Kipf and Max Welling. Semi-supervised classification with graph
convolutional networks. arXiv preprint arXiv:1609.02907, 2016.
[11] Yehuda Koren. Factor in the neighbors: Scalable and accurate collaborative
filtering. ACM Transactions on Knowledge Discovery from Data (TKDD),
4(1):1–24, 2010.
[12] Ping-Kan Liao and Von-Wun Soo. Neural networks to infer traditional chinese
medicine prescriptions from indications. In International Conference on Tech-
nologies and Applications of Artificial Intelligence, pages 198–216. Springer,
2023.
[13] Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Doll´ar.
Focal loss for dense object detection. In Proceedings of the IEEE international
conference on computer vision, pages 2980–2988, 2017.
[14] Yankai Lin, Zhiyuan Liu, Maosong Sun, Yang Liu, and Xuan Zhu. Learning
entity and relation embeddings for knowledge graph completion. In Proceed-
ings of the AAAI conference on artificial intelligence, volume 29, 2015.
[15] Steffen Rendle. Factorization machines. In 2010 IEEE International confer-
ence on data mining, pages 995–1000. IEEE, 2010.
[16] Keng-Chang Tsai, Yi-Chia Huang, Chia-Ching Liaw, Chia-I Tsai, Chun-
Tang Chiou, Chien-Jung Lin, Wen-Chi Wei, Sunny Jui-Shan Lin, Yu-Hwei
Tseng, Kuo-Ming Yeh, et al. A traditional chinese medicine formula nricm101
to target covid-19 through multiple pathways: A bedside-to-bench study.
Biomedicine & Pharmacotherapy, 133:111037, 2021.
[17] Yu-Hwei Tseng, Sunny Jui-Shan Lin, Sheng-Mou Hou, Chih-Hung Wang,
Shun-Ping Cheng, Kung-Yen Tseng, Ming-Yung Lee, Shen-Ming Lee, Yi-Chia
Huang, Chien-Jung Lin, et al. Curbing covid-19 progression and mortality
with traditional chinese medicine among hospitalized patients with covid-19:
a propensity score-matched analysis. Pharmacological research, 184:106412,
2022.
[18] Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero,
Pietro Lio, and Yoshua Bengio. Graph attention networks. arXiv preprint
arXiv:1710.10903, 2017.
[19] Hongwei Wang, Fuzheng Zhang, Jialin Wang, Miao Zhao, Wenjie Li, Xing Xie,
and Minyi Guo. Ripplenet: Propagating user preferences on the knowledge
graph for recommender systems. In Proceedings of the 27th ACM international
conference on information and knowledge management, pages 417–426, 2018.
[20] Hongwei Wang, Miao Zhao, Xing Xie, Wenjie Li, and Minyi Guo. Knowledge
graph convolutional networks for recommender systems. In The world wide
web conference, pages 3307–3313, 2019.
[21] Xiang Wang, Xiangnan He, Yixin Cao, Meng Liu, and Tat-Seng Chua. Kgat:
Knowledge graph attention network for recommendation. In Proceedings of
the 25th ACM SIGKDD international conference on knowledge discovery &
data mining, pages 950–958, 2019.
[22] Leihong Wu, Yi Wang, Zheng Li, Boli Zhang, Yiyu Cheng, and Xiaohui Fan.
Identifying roles of “jun-chen-zuo-shi” component herbs of qishenyiqi formula
in treating acute myocardial ischemia by network pharmacology. Chinese
medicine, 9:1–9, 2014.
[23] Yang Wu, Feilong Zhang, Kuo Yang, Shuangsang Fang, Dechao Bu, Hui Li,
Liang Sun, Hairuo Hu, Kuo Gao, Wei Wang, et al. Symmap: an integra-
tive database of traditional chinese medicine enhanced by symptom mapping.
Nucleic acids research, 47(D1):D1110–D1117, 2019.
[24] Liang Yao, Yin Zhang, Baogang Wei, Wenjin Zhang, and Zhe Jin. A topic
modeling approach for traditional chinese medicine prescriptions. IEEE
Transactions on Knowledge and Data Engineering, 30(6):1007–1021, 2018.
[25] William J Youden. Index for rating diagnostic tests. Cancer, 3(1):32–35,
1950.