论文部分内容阅读
目的应用径向基函数神经网络,在描述性分析的基础上,建立艾滋病病毒(HIV)感染者的预测模型,分析预测效果。方法在描述性分析的基础上,以患病种类为因变量,以年龄、国籍、文化程度、职业、劳务史、会阴部症状史、不洁性生活史、明确性伴侣有性病、是否有同性性伴侣、合法性伴侣是否是高危人群、性伴侣数为自变量,建立径向基函数神经网络,分析预测效果和预测变量的重要性。结果选择2004-2008年某省口岸出入境人员体检监测中检出的HIV感染者、梅毒患者和非性病人员各98例,HIV感染者以35～49岁男性为主(67.35%),中学以上文化程度占81.63%,职业以劳务(46.94%)和公务(30.61%)为主,有2例外籍女性性服务工作者和2例外教。径向基函数神经网络模型,对训练样本和检验样本的预测总准确率为85.37%和85.00%。用该模型对独立样本进行预测,总准确率为89.66%,对HIV感染者、梅毒患者和非性病患者的准确率分别为83.33%、85.71%和100.00%。三类预测结果ROC(Receiver operating characteristic)曲线下面积都>0.90,年龄、性伴侣数和劳务史是最重要的三个影响因素。结论口岸HIV感染者有其特定的流行病学特征,HIV感染者径向基函数神经网络预测模型的拟合能力强,能较好地用于评价未知样本。
Objective To establish a prediction model of HIV-infected persons based on descriptive analysis using radial basis function neural network and analyze the predictive effect. Methods On the basis of descriptive analysis, taking the prevalence of illness as the dependent variable and age, nationality, educational level, occupation, labor history, history of perineal symptoms, unclean sexual history, Sexual partners, whether legal partners are high-risk groups, the number of sexual partners as independent variables, the establishment of radial basis function neural network, analysis of the predictive effect and the importance of predictive variables. Results Ninety-eight cases of HIV-infected, syphilis and non-STD patients were detected in the surveillance of immigrants at a certain port in 2004-2008. Among HIV-infected persons, 35-49-year-old men were predominant (67.35%), Education level is 81.63%, occupation is labor service (46.94%) and official service (30.61%), there are 2 foreign female sex workers and 2 foreign teachers. Radial basis function neural network model, the training samples and test samples of the prediction accuracy of 85.37% and 85.00%. The accuracy of independent model was predicted by this model, with a total accuracy of 89.66%. The accuracy rates of HIV-infected, syphilis and non-STD patients were 83.33%, 85.71% and 100.00% respectively. The area under the receiver operating characteristic (ROC) curve of the three types of predictors is> 0.90. The age, number of sexual partners and service history are the most important three influencing factors. Conclusion The HIV-infected population in port has its own specific epidemiological characteristics. The prediction function of radial basis function neural network in HIV-infected population has good fitting ability and can be used to evaluate unknown samples.