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[2] 王世军[5] LIU Juan;ZHAO Huanhuan;MA Zuchang;WANG Shijun(38th Research Institute of China Electronics Technology Group Corporation,Hefei 230088;Hefei Institutes of Physical Science,Chinese Academy of Sciences,Hefei 230031;School of Science Island,University of Science and Technology of China,Hefei 230031;Chuzhou University,Chuzhou,Anhui Province 239000;5 Dalian Medical University,Dalian,Liaoning Province 116044)

机构地区：[1]中国电子科技集团公司第三十八研究所,合肥230088 [2]中国科学院合肥物质科学研究院,合肥230031 [3]中国科学技术大学科学岛分院,合肥230031 [4]滁州学院,安徽滁州239000 [5]大连医科大学,辽宁大连116044

出　　处：《北京生物医学工程》2020年第6期601-608,共8页Beijing Biomedical Engineering

基　　金：国家自然科学基金(61701482);安徽省自然科学基金(1808085MF191);安徽省科技重大专项(18030801133);中国科学院科技服务网络计划(KFJ-STS-ZDTP-079)资助。

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 3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4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 of 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 by using three machine learning algorithms,linear regression,artificial neural network and SVM,and were compared with the CatBoost model.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 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,and has better predictive capability compared with other traditional algorithms.Combined with fuzzy classification system,it can give users accurate prediction of blood pressure classification.

关 键 词：收缩压 舒张压 CatBoost 模糊分类系统 生活方式 

分 类 号：R318[医药卫生—生物医学工程]