北京生物医学工程

基于CatBoost算法和模糊分类系统的青年人血压预测方法

Blood pressure prediction method based on CatBoost algorithm and fuzzy classification system in young people

作者：刘娟赵欢欢马祖长王世军

单位：中国电子科技集团公司第三十八研究所（合肥230026）中国科学院合肥物质科学研究院（合肥230026）中国科学技术大学科学岛分院（合肥230026）滁州学院（安徽滁州239000）大连医科大学（辽宁大连 116044）

关键词：收缩压; 舒张压; CatBoost; 模糊分类系统; 生活方式

分类号：R318

出版年·卷·期（页码）：2020·39·6（601-608）

摘要：

目的探究CatBoost算法在青年人血压预测中的应用价值，为青年人高血压及高血压前期预警提供一种可行的技术手段。方法以2015-2017年期间在北京某医院体检中心进行健康体检的3872位青年人为研究对象，基于人口统计学和生活方式等指标,分别利用CatBoost算法构建收缩压预测模型和舒张压预测模型，然后利用模糊分类系统预测血压分级。使用线性回归、人工神经网络和SVM三种机器学习算法分别构建血压预测模型，并与CatBoost模型进行比较分析。以均方根误差（root mean square error，RMSE）和平均绝对百分比误差（mean absolute percentage error, MAPE）作为模型的评价指标。并进一步的分析模糊分类系统的预测效果。结果对于收缩压预测，基于CatBoost的模型在测试集上表现最优，RMSE和MAPE分别为11.17和7. 18%，对于舒张压预测，基于CatBoost的模型在测试集上表现最优，RMSE和MAPE分别为9.04和9.29%。进一步的模糊分类也取得了较好的血压分类准确性。变量重要性分析表明影响青年人血压值最重要的四个因素依次是年龄、身体质量指数(body mass index, BMI)、家族史和腰高比(waist-to-height ratio, WHtR)。结论 CatBoost算法在青年人血压预测中的应用具有一定的可行性，相比于其他传统算法，具有更好的预测能力。结合模糊分类系统，可以给用户较准确的血压分级预测。

Objective To explore the application value of CatBoost algorithm in blood pressure prediction for young people and provide a feasible technical means for early warning of hypertension and prehypertension in young people. Methods A total of 3872 young people who underwent a health check-up in a physical examination center in a Beijing hospital from 2015 to 2017 were selected as research objects, based on demography and lifestyle indicators, the CatBoost algorithm was used to construct a systolic blood pressure (SBP) prediction model and a diastolic blood pressure (DBP) prediction model respectively, and then fuzzy classification system was employed to predict blood pressure classification. In addition, blood pressure prediction models were constructed respectively using three machine learning algorithms, linear regression, artificial neural network and SVM, and compared with the CatBoost models. Root mean square error (RMSE) and mean absolute percentage error (MAPE) were used as the evaluation indicators of the models. The prediction effect of the fuzzy classification system was further analyzed. Results For SBP prediction, the CatBoost model performed the best on the test set, RMSE and MAPE were 11.17 and 7.18% respectively. For diastolic blood pressure prediction, the CatBoost model also performed the best on the test set, RMSE and MAPE were 9.04 and 9.29% respectively. Further fuzzy classification had also achieved good accuracy. The analysis of variable importance analysis showed that the four most important factors affecting blood pressure in young people were age, Body Mass Index(BMI), family history, and waist-to-height ratio(WHtR). Conclusions The CatBoost algorithm is feasible in the prediction of blood pressure in young people. Compared with other traditional algorithms, it has better predictive capability. Combined with fuzzy classification system, it can give users accurate prediction of blood pressure classification.

参考文献：

[1] Wu JH, Wei W, Zhang L, et al. Risk assessment of hypertension in steel workers based on LVQ and Fisher-SVM deep excavation[J]. IEEE Access, 2019, 7: 23109-23119.
[2] Ansah JP, Inn RLH, Ahmad S. An evaluation of the impact of aggressive hypertension, diabetes and smoking cessation management on CVD outcomes at the population level: a dynamic simulation analysis[J]. BMC Public Health, 2019, 19(1): 1105.
[3] Fernández-Ruiz I. Systolic and diastolic hypertension independently predict CVD risk[J]. Nature Reviews Cardiology, 2019,16(10): 578-579.
[4] Wang ZW, Chen Z, Zhang LF, et al. Status of hypertension in China results from the China Hypertension Survey, 2012-2015 [J]. Circulation, 2018, 137(22): 2344-2356.
[5] Ghosh S, Banerjee A, Ray N, et al. Continuous blood pressure prediction from pulse transit time using ECG and PPG signals[C]//IEEE Healthcare Innovation Point-of-Care Technologies Conference. Cancun, Mexico: IEEE Press, 2016.
[6] Munnoch R, Jiang P. A personal medical device for multi-sensor, remote vital signs collection in the elderly[C]// 2015 Science and Information Conference. London, UK: IEEE Press, 2015:1122-1131.
[7] Sideris C, Kalantarian H, Nemati E, et al. Building Continuous Arterial Blood Pressure Prediction Models Using Recurrent Networks[C]//2016 IEEE International Conference on Smart Computing (SMARTCOMP). St Louis, MO, USA: IEEE, 2016: 1-5.
[8] Simjanoska M, Kochev S, Tanevski J, et al. Multi-level information fusion for learning a blood pressure predictive model using sensor data [J]. Information Fusion, 2020, 58: 24-39.
[9] Kwong EW-Y, Wu H, Pang GK-H. A prediction model of blood pressure for telemedicine [J]. Health Informatics Journal, 2016，24（3）: 227-244.
[10] 徐继伟, 杨云. 集成学习方法:研究综述 [J]. 云南大学学报:自然科学版, 2018, 40(6): 36-46.
Xu JW, Yang Y. A survey of ensemble learning approaches[J]. Journal of Yunnan University（Natural Science）, 2018, 40(6): 36-46.
[11] 苗丰顺, 李岩, 高岑, 等. 基于CatBoost算法的糖尿病预测方法[J]. 计算机系统应用, 2019, 28(9): 215-218.
Miao FS, Li Y, Gao C, et al. Diabetes prediction method based on CatBoost algorithm[J]. Computer Systems & Applications, 2019, 28(9): 215-218.
[12] 王斌, 冯慧芬, 王芳, 等. 基于机器学习的Cat Boost模型在预测重症手足口病中的应用 [J]. 中国感染控制杂志, 2019, 18(1): 12-16.
Wang B, Feng HF, Wang F, et al. Application of CatBoost model based on machine learning in predicting severe hand- foot-mouth disease[J]. Chinese Journal of Infection Control, 2019, 18(1): 12-16.
[13] 《中国高血压防治指南》修订委员会. 中国高血压防治指南2018年修订版 [J]. 心脑血管病防治, 2019, 19(1): 1-44.
[14] 张冬冬, 刘雪娇, 王炳源, 等. 我国成年人群年龄对体质指数与高血压发病风险的修饰效应[J]. 中华流行病学杂志, 2018, 39(6): 765-769.
Zhang DD, Liu XJ, Wang BY, et al. Age-related modification effect on the association between body mass index and the risk of hypertension: A Cohort Study on Chinese people living in the rural areas [J]. Chinese Journal of Epidemiology, 2018, 39(6): 765-769.
[15] 李婧雯, 张晓卉, 尹新华. 肥胖相关高血压的研究进展[J]. 临床与病理杂志, 2020, 40(4): 1006-1011.
Li JW, Zhang XH, Yin XH. Progress in obesity-related hypertension[J]. Journal of Clinical and Pathological Research, 2020, 40(4): 1006-1011.
[16] Zhang M, Zhao Y, Wang G, et al. Body mass index and waist circumference combined predicts obesity-related hypertension better than either alone in a rural Chinese population [J]. Scientific Reports, 2016, 6： 31935.
[17] Huang XB, Chen F, Dai W, et al. Prevalence and risk factors associated with hypertension in the Chinese Qiang population[J]. Clinical and Experimental Hypertension, 2018, 40(5): 427-433.
[18] 张艳艳, 何朝, 赵莹颖, 等. 北京市顺义区人群中心型肥胖与心血管疾病危险因素分析[J]. 中国健康教育, 2016 , 32( 5) : 419-423.
Zhang YY, He C, Zhao YY, et al. Relationship between central obesity and cardiovascular risk factors among residents of Shunyi District, Beijing[J]. Chinese Journal of Health Education, 2016 , 32( 5) : 419-423.
[19] 苏健,向全永,吕淑荣,等.江苏省成人中心型肥胖与心血管疾病危险因素及其聚集性的关系[J].中华心血管病杂志,2015,43( 6) : 548-553.
Su J, Xiang QY, Lyu SR, et al. Relationship between central obesity and clustering of cardiovascular risk factors in adults of Jiangsu Province [J] . Chinese Journal of Cardiology, 2015,43( 6) : 548-553.
[20] 沈振海, 陆昀, 祁华金,等. 理想心血管健康与腰围、腰高比的相关性研究[J]. 中华健康管理学杂志, 2017, 11(1): 56-61.
Shen ZH, Lu Y, Qi HJ, et al. Association between ideal cardiovascular health and waist circumference, waist-to-height ratio [J]. Chinese Journal of Health Management, 2017, 11(1): 56-61.
[21] 汪宏莉, 韩延柏, 陈涛, 等. 老年人高血压风险预测肥胖指标筛选[J]. 中国公共卫生, 2020, 36(1): 105-109.
Wang HL, Han YB, Chen T, et al. Screening on obesity indexes for predicting hypertension risk in 60 – 69 years old community residents[J]. Chinese Journal of Public Health, 2020, 36(1): 105-109.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录