摘要
目的建立皖南山区流行性腮腺炎(简称流腮)发病率智能预测模型,以提高预警能力。方法借助Matlab6.5软件中的神经网络工具箱,以安徽省池州市2005-2008年间的月平均气温、月平均相对湿度、月日照时间、月疫苗使用量和上月流腮发病率5个指标作为输入向量,同期流腮月发病率为目标向量,利用优化的levenberg-marquardt算法建立BP神经网络预测模型,并对模型进行验证。结果以2009年1月～2010年4月数据为测试样本提供给已建立的5×11×1结构的BP网络模型进行仿真预测,与实际发病率很好吻合,其误差均衡地分布在0值附近,表明该模型预测能力良好。结论利用BP网络进行疾病趋势预测,可获得更好的预测效果,且对资料的类型和分布不作任何限制,是一种全新的流行病学预测方法,其应用前景广阔。
Objective To explore a intellectualized forecast model on the incidence of mumps in south- ern Anhui Province for the purpose of getting warning signals beforehand. Methods By using Matlab 6.5, one back-propagation Neural Network （BPNN） was established by levenberg-marquardt method with 5 indexs of Chizhou City （average monthly air-temperature, relative humidity, sunshine-hours, monthly vaccine-consumption and the incidence of mumps last month） among 2005 - 2008 as input-vector and the monthly incidence of mumps of the same term as output-vector. The established BPNN model was tested as well. Results The forecastting values from BPNN fit the actual incidence of mumps well from Jan. 2009 to Apr. 2010 with its errors between forecast and actuality scatterting evenly around 0. It makes clear that the established BPNN model＇ with 5 × 11 × 1 framework displayed high forecast capability on the incidence of mumps in southern Anhui Province. Conclusions Compared to the conventional statistic method, BPNN model not only showed better prediction precision, but had no limit to the type or distribution of relevant data. Thus it provided us a brand-new powerful method in epidemiological prediction with a predictable wider application.
出处
《中华疾病控制杂志》
CAS
2010年第8期739-741,共3页
Chinese Journal of Disease Control & Prevention
