| 引用本文: |
孙资金,吉静,马重阳,张风君,赵宏跃,王雪茜,王庆国,程发峰.基于机器学习的中风中医辨证模型的构建与应用[J].湖南中医药大学学报,2023,43(4):694-699[点击复制] |
|
| |
|
|
| 本文已被：浏览 2671次 下载 1638次 |
| 基于机器学习的中风中医辨证模型的构建与应用 |
| 孙资金,吉静,马重阳,张风君,赵宏跃,王雪茜,王庆国,程发峰 |
| (北京中医药大学中医学院, 北京 100029;首都医科大学中医药学院, 北京 100069;山东中医药大学针灸推拿学院, 山东 济南 250355;哈尔滨医科大学附属第一医院, 黑龙江 哈尔滨 150007) |
| 摘要: |
| 目的 建立基于人工智能的中风中医辨证模型，为中风中医智能辨证模型的构建与应用提供方法和依据。方法 检索中国期刊全文数据库，收集关于中风的中医病案五种证型各60例，建立中风病案中医信息数据库，采用经过超参数调优的支持向量机（support vector machine, SVM）、K-近邻（K-nearest neighbor, KNN）、随机森林（random forest, RF）、极端随机树（extremely randomized trees, ExtraTrees）、XGBoost及LightGBM对数据进行机器学习建模。全部数据的70%作为训练集，30%作为测试集，采用五折交叉验证对模型进行评价，以Accuracy作为模型优劣的评价指标，比较模型的准确性。结果 中风中医四诊信息为输入变量共55项，中风中医证型为输出变量共5项。6种模型的拟合效果较好，Accuracy值均在0.85以上；其中SVM模型的准确率最高，可达0.95。结论 基于SVM算法模型建立的中风中医辨证模型具有较好的诊断、预测能力，机器学习技术应用于中风中医辨证模型的构建具有方法学上的可行性。 |
| 关键词: 人工智能 机器学习 中风 中医辨证模型 中医药现代化 证候 大数据 |
| DOI：10.3969/j.issn.1674-070X.2023.04.019 |
| 投稿时间：2022-09-13 |
| 基金项目:国家自然科学基金项目（U21A20400）；燕京刘氏伤寒流派传承工作室项目（1190062620029）。 |
|
| Construction and application of stroke TCM pattern differentiation model based on machine learning |
| SUN Zijin,JI Jing,MA Chongyang,ZHANG Fengjun,ZHAO Hongyue,WANG Xuexi,WANG Qingguo,CHENG Fafeng |
| (School of Chinese Medicine, Beijing University of Chinese Medicine, Beijing 100029, China;School of Chinese Medicine, Capital Medical University, Beijing 100069, China;School of Acupuncture-Moxibustion and Tuina, Shandong University of Chinese Medicine, Jinan, Shandong 250355, China;The First Hospital of Harbin Medical University, Harbin, Heilongjiang 150007, China) |
| Abstract: |
| Objective To establish a TCM pattern differentiation model of stroke based on artificial intelligence, and to provide methods and basis for the construction and application of TCM intelligent pattern differentiation model of stroke.Methods Chinese Journal Full-text Database (CJFD) was searched for the five pattern types of TCM medical records regarding stroke, with 60 cases in each type. Then a TCM information database of stroke medical records was established. Support vector machine (SVM), K-nearest neighbor (KNN), random forest (RF), extreme random trees (Extra Trees), XGBoost, and LightGBM after hyper-parameter optimization were used to construct machine learning models, and 70% of the total data was used as the training set and 30% as the test set. Meanwhile, five-fold cross-validation was used to evaluate each model, and Accuracy was the evaluation index to compare the models accuracy.Results There were 55 input variables (information of stroke obtained by four diagnostic methods of TCM), and 5 output variables (TCM patterns of stroke). The fitting effect of the six models was good, and the accuracy values were all above 0.85; among which, the accuracy of SVM model was the highest, up to 0.95.Conclusion The TCM pattern differentiation model of stroke based on the SVM algorithm can diagnose and predict well, therefore, it is methodologically feasible to apply machine learning technology for constructing TCM pattern differentiation model of stroke. |
| Key words: artificial intelligence machine learning stroke TCM pattern differentiation model TCM modernization pattern big data |
|
二维码（扫一下试试看！） |
|
|
|
|
