Selection of Machine Learning Models for Prescription Decision-Making Based on Text Mining - Focusing on Case Studies of Single Prescriptions in Sasang Constitutional Medicine

Kim Yunseo; Lee Eunsu; Jeong Roa; Park Sohyun; Kim Jinseok; Kim Hyeonseo; Yu Jun-sang; Kim, Yunseo; Lee, Eunsu; Jeong, Roa; Park, Sohyun; Kim, Jinseok; Kim, Hyeonseo; Yu, Jun-sang

doi:10.13048/jkm.25006

JKM > Volume 46(1); 2025 > Article

Yunseo, Eunsu, Roa, Sohyun, Jinseok, Hyeonseo, and Jun-sang: Selection of Machine Learning Models for Prescription Decision-Making Based on Text Mining - Focusing on Case Studies of Single Prescriptions in Sasang Constitutional Medicine

Original Article

The Journal of Korean Medicine 2025; 46(1): 70-86.

Published online: March 01, 2025

DOI: https://doi.org/10.13048/jkm.25006

Selection of Machine Learning Models for Prescription Decision-Making Based on Text Mining - Focusing on Case Studies of Single Prescriptions in Sasang Constitutional Medicine

Kim Yunseo¹^{, §}

, Lee Eunsu¹^{, §}

, Jeong Roa¹^{, §}

, Park Sohyun¹

, Kim Jinseok¹

, Kim Hyeonseo²

, Yu Jun-sang³^{, 4}^{, *}

¹Department of Korean Medicine, Sangji University

²Department of Korean Language & Literature, Sogang University

³Department of Sasang Constitutional Medicine, College of Korean Medicine, Sangji University

⁴Research Institute of Korean Medicine, Sangji University

Correspondence to: Jun-sang Yu, Korean Medicine Hospital of Sangji University, 80 Sangjidae-gil, Wonju-si, Gangwon-do, 26338, Republic of Korea, Tel: +82-33-741-9203, E-mail: hiruok@sangji.ac.kr

§ These authors contributed equally to this work.

Received January 20, 2025 Revised January 20, 2025 Accepted February 11, 2025

Abstract

Objectives

We analyzed Sasang constitution case reports using text mining and designed a classification algorithm using machine learning to select a model suitable for determining Sasang constitution prescriptions based on text data.

Methods

Case reports on Sasang constitution published from January 1, 2000, to December 31, 2023, were collected. A total of 360 papers and 483 cases were identified, from which text was extracted for 253 cases. The extracted texts were preprocessed and tokenized using the Python-based KoNLPy package, and each morpheme was vectorized using TF-IDF values. To select the most suitable classification model for diagnosing Sasang constitution, the performance of five models—Random Forest Classifier, XGBoost, LightGBM, SVM, and Logistic Regression—was evaluated based on accuracy and F1-Score.

Results

The highest accuracy was achieved by Random Forest Classifier (0.57037), followed by SVM (0.544444), Logistic Regression (0.518519), LightGBM (0.481481), and XGBoost (0.474074). The F1 score was highest for Random Forest Classifier (0.528), followed by SVM (0.52039), Logistic Regression (0.500861), XGBoost (0.45866), and LightGBM (0.446349).

Conclusions

This study is the first to analyze Sasang constitution prescription decisions by applying text mining and machine learning to case reports, providing a concrete research model for follow-up studies. Based on case reports and text data, the most suitable machine learning model for determining Sasang constitution prescriptions is Random Forest Classifier.

Keywords: Data Mining, Machine Learning, Case reports, Four-Constitution Medicine, Random Forest Classifier

\begin{matrix} t f - i d f (t, d) = t f (t, d) \times i d f (t) \\ i d f (t) = log \frac{1 + n_{d}}{1 + d f (d, t)} + 1 \end{matrix}

Fig. 1

Study flow of Text-mining and Machine learning

Fig. 2

Flow chart of literature searches and screening results

Fig. 3

Performance Comparison of Different Algorithms

Table 1

Data Refining Criteria

Criteria		Example
Criteria		Before	After
Compound words	Cases where a compound word is perceived as separate components	Cold, Sweat	Cold sweat
		Hyung, Geumji, Pose	Hyunggeumjipose
	Cases where multiple words should be considered as a single phrase	Abdominal, Bloating	Abdominal bloating
		Nocturnal, sleep, disorder	Nocturnal sleep disorder
Synonyms	Cases with the same or similar meanings but different spellings	Feel dizzy, Dizziness, Lightheadedness	Vertigo
Synonyms	Cases where a single word represents or encompasses other words	Sleep disorder, Difficulty falling asleep, Nocturnal sleep disorder, Insomnia, Sleep difficulties, Difficulty falling asleep	Sleep disorder
Stop words	Not a key variable, and used conventionally	Above-mentioned, Opinion, Usually, And, When, Time, Patient	Delete

Table 2

English Word Translation Exclusion Criteria

Translation exclusion criteria	Example
Words written in English in most of the research papers	VAS, QSCC II
Words that represents a unit	kg, cm
Name of the medicine	trolac, NSAID

Table 3

Hyperparameter Settings

Algorithm	Hyper Parameter	Input Value
Random Forest Classifier	n_estimators	50, 200, 500, 1000

XGBoost	learning_rate	0.01, 0.1, 0.2
XGBoost	n_estimators	100, 200

LightGBM	learning_rate	0.01, 0.1, 0.2
LightGBM	n_estimators	100, 200

Support Vector Machine	C	0.1, 1, 10, 20
	kernel	‘linear’, ‘rbf’, ‘sigmoid’, ‘poly’
	degree	2, 3, 4

Logistic Regression	C	0.1, 1, 10, 20

Table 4

Best Hyperparameter Settings

Algorithm	Hyper Parameter	Best Hyper Parameter	Best CV F1 Score
Random Forest Classifier	n_estimators	500	0.508679386

XGBoost	learning_rate	0.1	0.490174846
XGBoost	n_estimators	200	0.490174846

LightGBM	learning_rate	0.1	0.456298434
LightGBM	n_estimators	200	0.456298434

Support Vector Machine	C	10	0.484772635
	kernel	‘sigmoid’
	degree	2

Logistic Regression	C	20	0.482295398

Table 5

Number of Cases Based on Use of Single Prescription by Sasang Constitution

	Number of case studies with a single prescription	Number of case studies with compound prescriptions	Total number of case studies
Taeyangin	14	4	18
Taeeumin	73	81	154
Soyangin	111	105	216
Soeumin	55	40	95
Total	253	230	483

Table 6

Frequency of Prescriptions by Sasang Constitution (Single Prescription)

Taeyangin		Taeeumin		Soyangin		Soeumin
Mihudeungsikjang-tang	11	Yeoldahanso-tang	18	Yanggyeoksanhwa-tang	21	Gwakhyangjeonggi-san	14
Ogapijangcheok-tang	2	Cheongsimyeonja-tang	16	Hyeongbangjihwang-tang	18	Sibimigwanjung-tang	10
Yeoldahanso-tang	1	Taeumjowi-tang	9	Hyeongbangdojeok-san	18	Hyangbujapalmul-tang	3
		Jowiseungcheong-tang	8	Hyeongbangsabaek-san	17	Palmulgunja-tang	3
		Galgeunhaegi-tang	7	Dojeokganggi-tang	7	Seungyangikgibuja-tang	3
		Gwache	3	Dokhwaljihwang-tang	7	Geopung-san	2
		Joripyeowon-tang	2	Yangdokbaekho-tang	6	Hyangsayangwi-tang	2
		Cheongpyesagan-tang	2	Gamsumal	5	Cheongunggyeji-tang	2
		Geonyuljeotang-tang	1	Hyeongbangpaedok-san	4	Osuyubujaijung-tang	2
		Mahwangjeongcheon-tang	1	Yukmijihwang-tang	1	Doksampalmulgunja-tang	2
		Mankgeummunmu-tang	1	Jeoryeongchajeon-tang	1	Bojungikgi-tang	2
		Seunggeumjowi-tang	1	Jihwangbaekho-tang	1	Oryeong-san	1
		Seunggijowi-tang	1	Palmulgunja-tang	1	Seonghyangjeonggi-san	1
		Cheongrijagam-tang	1	Sukjihwanggosam-tang	1	Seungyangikgi-tang	1
		Cheonghyeolgangih-tang	1	Saenghwa-tang	1	Samgyepalmul-tang	1
		Handayeolso-tang	1	Ganghwajihwang-tang	1	Doksamgwangyebujaijung-tang	1
				Gamijihwang-tang	1	Dangguibaekhaoogwanjung-tang	1
						unggihyangso-san	1
						Gunggichiseup-tang	1
						Gwangyebujaijung-tang	1
						Hwanggyeogyeji-tang	1
	14		73		111		55

Table 7

Average Accuracy, F1-score, Precision, and Recall of Algorithms

Algorithm	Average Accuracy	Average F1 score	Average Precision	Average Recall
Random Forest Classifier	0.57037	0.528	0.548862	0.557407
XGBoost	0.474074	0.45866	0.477778	0.482407
LightGBM	0.481481	0.446349	0.454074	0.478704
Support Vector Machine	0.544444	0.52039	0.548003	0.546296
Logistic Regression	0.518519	0.500861	0.533915	0.52037

참고문헌

1. Chang, JY. (2013). A Study on Research Trends of Graph-Based Text Representations for Text Mining. The Journal of the Institute of Internet Broadcasting and Communication, 13(5), 37-47. https://doi.org/10.7236/JIIBC.2013.13.5.37

2. Koo, HI. (2018). AI and Deep Learning Trends. The Korean Institute of Electrical Engineers, 67(7), 7-12.

3. Kim, NG., Lee, DH., Choi, HC., & Wong, WXS. (2017). Investigations on Techniques and Applications of Text Analytics. The Journal of Korean Institute of Communications and Information Sciences, 42(2), 471-492. https://doi.org/10.7840/kics.2017.42.2.471

4. Eum, SW. (2020). A Study on Analysis of consumer perception of YouTube advertising using text mining. Management & Information Systems Review, 39(2), 181-193. https://doi.org/10.29214/DAMIS.2020.39.2.011

5. Jung, M., Lee, YL., Yoo, CM., Kim, JW., & Chung, JE. (2019). An exploratory study on consumers’ responses to mobile payment service focused on Samsung Pay. Journal of Digital Convergence, 17(1), 9-27. https://doi.org/10.14400/JDC.2019.17.1.009

6. Choi, HJ. (2022). Comparison of Machine Learning Methods for a Prediction of Match Outcomes in Soccer. The Journal of the Korean Society of Measurement and Evaluation in Physical Education and Sports Sceince, 24(4), 81-91. http://doi.org/10.21797/ksme.2022.24.4.081

7. Park, HS., Lee, MS., Hwang, SJ., & Oh, SY. (2016). TF-IDF Based Association Rule Analysis System for Medical Data. The Transactions of the Korea Information Processing Society, 5(3), 145-154. https://doi.org/10.3745/KTSDE.2016.5.3.145

8. Cho, SZ., & Kang, SH. (2016). Industrial Applications of Machine Learning (Artificial Intelligence). Industrial Engineering Magazine, 23(2), 34-38.

9. Jang, DY., Ha, YS., Lee, CY., & Kim, CE. (2020). Analysis of Symptoms-Herbs Relationships in Shanghanlun Using Text Mining Approach. Journal of Physiology & Pathology in Korean Medicine, 34(4), 159-169. https://doi.org/10.15188/kjopp.2020.08.34.4.159

10. Bae, HJ., Kim, CE., Lee, CY., Shin, SW., & Kim, JH. (2018). Investigation of the Possibility of Research on Medical Classics Applying Text Mining - Focusing on the Huangdi’s Internal Classic -. Journal of Korean Medical classics, 31(4), 27-46. https://doi.org/10.14369/jkmc.2018.31.4.027

11. Yea, SJ. (2023). Analysis of Papers on Side-Effects Caused by Herbal Medicine Prescription Using Text Mining: Leveraging PubMed Articles. Journal of Knowledge Information Technology and Systems, 18(3), 501-511. https://doi.org/10.34163/jkits.2023.18.3.001

12. Yea, SJ., & Kim, SH. (2022). An Analysis of the Research Trends of Five Traditional Korean Medicine Prescriptions Using Text Mining: Leveraging PubMed Articles. Journal of Knowledge Information Technology and Systems, 17(5), 815-823. https://doi.org/10.34163/jkits.2022.17.5.003

13. Kim, JS., Park, SH., Jeong, RA., Lee, ES., Kim, YS., Sung, HD., & Yu, JS. (2024). Application of text-mining technique and machine-learning model with clinical text data obtained from case reports for Sasang constitution diagnosis: a feasibility study. The Journal of Korean Medicine, 45(3), 193-210. http://dx.doi.org/10.13048/jkm.24049

14. Park, MS., Kim, MH., Park, SY., Choi, IH., & Kim, CE. (2022). Individualized Diagnosis and Prescription in Traditional Medicine: Decision-Making Process Analysis and Machine Learning-Based Analysis Tool Development. The American Journal of Chinese Medicine, 50(7), 1827-1844. https://doi.org/10.1142/S0192415X2250077X

15. Cho, IH., Kwon, JH., Lee, EJ., & Lee, JH. (2020). A Study on Clinical Status for Development of Clinical Practice Guidelines for Sasang Constitutional Medicine Symptomatology. Journal of Sasang Constitution and Immune Medicine, 32(4), 29-44. https://doi.org/10.7730/JSCM.2020.32.4.29

16. Jeon, TH. (2022). A linguistic study on tokenization methods for Korean text. Language Facts and Perspectives, 55, 309-354. https://doi.org/10.20988/lfp.2022.55..309

17. Park, HJ. (2020). Trend Analysis of Korea Papers in the Fields of ‘Artificial Intelligence’, ‘Machine Learning’ and ‘Deep Learning. Journal of Korea Institute of Information, Electronics, and Communication Technology, 13, 283-292. http://doi.org/10.17661/jkiiect.2020.13.4.283

18. Lee, JH., Lee, MB., & Kim, JW. (2019). A study on Korean language processing using TF-IDF. The Journal of Information Systems, 28(3), 105-121. http://dx.doi.org/10.5859/KAIS.2019.28.3.105

19. Park SE., Gang JY.(2022). Python Text Mining Complete Guide. 1st Edition. Gyeonggi. Wikibooks;p. 322

20. Hong, KH. (2020). A Predictive Model for Suicidal Ideation of Adolescents Using Random Forests Machine Learning Algorithm. Korean Journal of Social Welfare, 72(3), 157-180. https://doi.org/10.20970/kasw.2020.72.3.007)

21. Bae, JS., & Kim, SB. (2021). Predictions of COVID-19 in Korea Using Machine Learning Models. Journal of the Korean Institute of Industrial Engineers, 47(3), 272-279. https://doi.org/10.7232/JKIIE.2021.47.3.272

22. Hah, DW., Kim, YM., & Ahn, JJ. (2019). A study on KOSPI 200 direction forecasting using XGBoost model. Journal of the Korean Data And Information Science Sociaty, 30(3), 655-669. http://dx.doi.org/10.7465/jkdi.2019.30.3.655

23. Hwang, YJ., Son, SE., & Lee, ZK. (2024). Prediction of Stock Returns from News Article’s Recommended Stocks Using XGBoost and LightGBM Models. Journal of The Korea Society of Computer and Information, 29(2), 51-59. http://doi.org/10.9708/jksci.2024.29.02.051

24. Park, SY., & Chung, HW. (2020). Exploring Variables Affecting Career Decision of Middle School Students: An Application of Machine Learning Approaches. Asian Journal of Education, 21(3), 727-753. http://doi.org/10.15753/aje.2020.09.21.3.727

25. Kim, PS., & Lee, SH. (2023). Application of AI Machine Learning Algorithms to Predict Korea Ladies Professional Golf Association (KLPGA) Players Top 10 Ranking: A Sports Analytics Perspective. Korean Journal of Sport Management, 28(4), 51-66. http://doi.org/10.31308/KSSM.28.4.51

26. Jung, MH., & Kwon, WH. (2021). Present Status and Future of AI-based Drug Discovery. Journal of the Korea Institute Of Information and Communication Engineering, 25(12), 1797-1808. http://doi.org/10.6109/jkiice.2021.25.12.1797

27. Lee JH.(2022). Korean Medicine Clinical Practice Guideline for Sasang(Four) constitutional medicine patterns. Korea. The Society of Sasang Constitutional Medicine.

28. National Institute of Korean Medicine Development. 2017 Korean Medicine Health Service Utilization and Consumption Survey. Seoul. (2018.

29. Lee, JH., & Lee, HH. (2019). Selecting Sasang-Type classification model using machine learning and designing the service flow. Journal of Digital Contents Society, 20(2), 321-327. http://dx.doi.org/10.9728/dcs.2019.20.2.321

30. Rácz, A., Bajusz, D., & Héberger, K. (2021). Effect of Dataset Size and Train/Test Split Ratios in QSAR/QSPR Multiclass Classification. Molecules, 26(4), 1111. https://doi.org/10.3390/molecules26041111

31. Oh, SM., & Park, MS. (2024). Machine Learning-based Phishing Website Detection Model. The Journal of the Convergence on Culture Technology, 10(4), 575-580. http://dx.doi.org/10.17703/JCCT.2024.10.4.575

32. Han, KS., & Park, SS. (2002). Research about The Discourse on The Discourse on The Medications and Prescriptions on The ShinChukBon DongyiSuseBowon. Journal of Sasang Constitution and Immune Medicine, 14(3), 52-73.

33. Department of Sasang Constitutional Medicine, College of Korean Medicine. (2004). Sasang constitutional medicine. Jipmoon, 53.

34. Kang, MS., Oh, JW., Lee, HR., & Lee, JH. (2019). Patient Group Study to Improve the Accuracy of QSCC II+. Journal of Sasang Constitution and Immune Medicine, 31(3), 48-65. http://doi.org/10.7730/JSCM.2019.31.3.48