用人工神经网络方法建立出入境人员性病艾滋病预测模型

Establishing STD/AIDS Forecasting Model with the Artificial Neural Network among Entry-exit Personnels

〔目的〕研究用人工神经网络建立出入境人员性病艾滋病预测模型。〔方法〕以330例患有性病、艾滋病的出入境人员和330例非性病出入境人员组成研究样本。在对研究对象问询资料整理后进行统计分析和统计推断,再运用BP神经网络(BP-ANN)拟合,以15个自变量(包括年龄、性别、国籍、职业、文化程度、国内外劳务史、性伴侣数、生殖系统异常症状史、不洁性生活史、明确性伴侣是否是有性病、合法性伴侣是否性病的高危人群、输血史、吸毒史、是否有同性性伴侣和是否拒绝流行病学调查)建立对患病的预测模型。〔结果〕性病、艾滋病患者与非性病者间在年龄、性别、国籍、职业、文化程度、国内外劳务史、性伴侣数、病史、不洁性生活史、性伴方面的差异都具有显著性(P<0.01)。以隐含层神经元个数为7个BP-ANN模型中的预测效果最好,网络的结构为15-7-1。该模型的准确率较高,训练准确率、校验准确率和测试准确率分别达到93.94%、88.48%和89.60%,同时稳定性高、复杂性低。〔结论〕BP-ANN较好的建立了性病、艾滋病预测模型,为口岸和国内外性病、艾滋病监测和风险预警提供了一种新的方法。

Objective To establish STD/AIDS＇s predictive model by using artificial neural network. Methods 330 entry-exit personnels who were not patients of STD and 330 entry-exit patients of STD （including AIDS） were studied.While the data were collected and analyzed,the BP-ANN was set up by the independent variables included age,sex,nationality,occupation,education,labor history,number of sexual partner, abnormal symptom of reporoductive system,risk sexual behaviour, whether sexual partner had STD, whether his/her spouse was a STD patient,history of blood transfusion,history of drug use,whether having homogeneity sexual partner, whether refusing epidemiologic survey. Results There were insignificant differences between the STD patients and not STD patients in age,sex,nationality,occupation,education,labor history,number of sexual partner, history of abnormal symptom of reporoductive system,risk serual behaviour,whether sexual partner was STD patient,whether his/her spouse was a STD patient,having history of blood transfusion, history of drug user,whether having homogeneity sexual partner,Refusing epidemiologic survey （P〈 0.01）.The BP-ANN which had 7 implying neurone was the best one. It was accurate, stable and low complexity. The accuracy for training, checking and testing were 93.94%,88.48% and 89.60% respectively. Conclusion BP-ANN will set up a good model for forecasting STD/AIDS, and provide a new method for STD/AIDS＇s monitoring, also provide a new technology for risk precaution.