摘要
为了帮助医疗机构合理调配医务力量、床位和医疗药物,同时也帮助脑卒中高危人群及时采取干预措施,降低发病风险。本文对某市[1]四家医院2013—2016年脑卒中的就诊病例进行数据分析,将日就诊人数分为6个等级。然后,调取相应时段当地的逐日气象资料,采用支持向量机(SVM)和随机森林(RF)方法分别建立了日就诊人数预测模型和日就诊人数与气象因素的关系模型。研究结果表明:(1)脑卒中的日就诊人数为不平衡数据,这种数据特征将导致传统的预测模型正确率较低;(2)通过不断调整SVM预测模型的初始权重,经历了4次优化之后,使得日就诊人数的预测正确率从52.46%上升到94.56%;(3)随机森林模型的结果显示,影响脑卒中发病率的三大气象因素分别是最高气温、最低气温和平均气温。基于机器学习模型的脑卒中疾病与气象因素的研究成果,提高了医疗气象统计模型的预报准确率,具有较高的应用和推广价值。
The purpose of this paper is to help medical institutions in terms of a reasonable allocation of medical hands, beds and medical drugs, and also to help people in high stroke risk by mean of timely inter- vention, reducing the risk of stroke occurrence. The data of the cases of stroke in four hospitals from 2013 to 2016 were analyzed, and the number of daily patients in the hospital was divided into six grades. By tak- ing the support vector machine （SVM） and the random forest （RF） methods, and using the corresponding local daily meteorological data collected, we established the prediction model for daily patients and the cor- relation model related to meteorological factors. The results showed that： （1） the number of daily patients with stroke was not balanced and the accuracy of traditional prediction model was relatively low. （2） By constantly adjusting the initial weight of the SVM prediction model, we found that, after 4 attempts of op- timization, the prediction accuracy of the number of daily patients increased from 52.46% to 94.56%. （3） The results of the random forest model showed that the three meteorological factors affecting the incidence of stroke are the maximum temperature, the minimum temperature and the mean temperature. Thus, the research results of the correlation between stroke diseases and meteorological factors based on machine learning model could improve the prediction accuracy of medical meteorological statistical models, and have good application and promotion values.
作者
程学伟
韩兆洲
CHENG Xuewei;HAN Zhaozhou(Department of Statistics, College of Economics, Jinan University, Guangzhou 510632)
出处
《气象》
CSCD
北大核心
2018年第6期837-843,共7页
Meteorological Monthly
基金
国家社会科学重点基金项目(15ATJ001)资助
关键词
脑卒中
气象因素
SVM
RF
非平衡数据
stroke
meteorological factors
support vector mechine （SVM）
random forest （RF）
nonequilibrium data
作者简介
程学伟，主要从事机器学习算法以及气象统计研究．Email：cxwpaper@163．com
