基于AdaBoost和分类树的北京市高血压患者就诊机构选择的影响因素分析

发布时间：2018-03-30 16:14

本文选题：AdaBoost　切入点：分类树　出处：《中国卫生统计》2017年01期

【摘要】：目的通过对2013年"北京市第五次国家卫生服务调查"中高血压患者两周就诊数据的分析,探讨高血压患者两周就诊机构的选择及影响因素。方法使用Ada Boost和分类树组合分类器对高血压患者两周首次就诊机构进行分类,获得变量相对重要性排序结果,利用十折交叉验证法计算错分率,错分率的可接受程度为0~0.3,选择错分率最小的模型进行结果解释。结果本次1128例患者中,两周首次就诊选择基层医疗卫生机构的占75.7%。选用的AdaBoost和分类树组合分类器的模型错分率为0.177。经Ada Boost和分类树组合分类器对10个自变量按照相对重要性进行排序,其中排名最高的三个变量依次为受教育程度、家庭人均收入、家人常去医疗机构,并绘制能够分别说明这三个变量的分类树。结论对于北京市两周就诊的高血压病患者来说,首诊选择基层医疗卫生机构的比例较高,受教育程度、家庭人均收入和家人常去医疗机构是影响其是否选择基层医疗卫生机构的最重要因素。
[Abstract]:Objective to analyze the two-week visit data of hypertension patients in the Fifth National Health Service Survey of Beijing in 2013. Methods Ada Boost and classification tree combined classifier were used to classify the first visit organization in two weeks of hypertension patients, and the results of relative importance ranking of variables were obtained. Ten fold cross-validation method was used to calculate the error score rate. The acceptable degree of error score rate was 0 / 0.3. The model with the lowest error score rate was selected to explain the results. Results of the 1128 patients, In two weeks, 75.75.75% of the primary medical and health institutions were selected for the first visit. The model error rate of the selected AdaBoost and classification tree combination classifier was 0.177. Ten independent variables were sorted according to relative importance by Ada Boost and classification tree combination classifier. The top three variables in order are education level, family income per capita, family members often go to medical institutions, and draw the classification tree which can explain these three variables separately. Conclusion for the patients with hypertension in Beijing two weeks, The proportion of primary medical and health institutions is high, the education level, the average family income and the family often go to medical institutions are the most important factors that influence the selection of primary health care institutions.
【作者单位】：北京大学公共卫生学院;北京市公共卫生信息中心;郑州大学第一附属医院;
【分类号】：R544.1

【相似文献】