基于时间序列的传染病预测应用研究

【作者】 李鹏；

【导师】 韩晓东；

【作者基本信息】 昆明理工大学 ， 计算机技术（专业学位）， 2018， 硕士

【摘要】 传染病对人们的生活造成了极大的危害,轻则会造成患者身体的不适,重则会严重影响患者身体各个器官的功能障碍,甚至危及生命。因此,对传染病的发病率进行较为准确预测就显得十分重要和迫切。对传染病的发病率进行预测不仅可以为相关部门提供参考、依据,使相关部门提前作出正确的决策,而且可以提高公众的传染病危险性意识,使传染病在某种程度得到预防和控制。本文使用时间序列预测的方法对云南省乙类传染的发病率进行分析和研究,通过BP神经网络模型、ARIMA模型对云南省乙类传染病的发病率进行预测,最后,把BP神经网络模型和ARIMA模型结合起来,形成了新的组合模型。本论文主要的工作如下:第一,根据云南省年度人口数据建立灰色预测模型,用2005-2016年的数据预测2017、2018年的云南省年度总人口数据。结果显示,模型的预测等级为一级,表明该模型可以很好的用在云南省人口数据预测上。第二,云南省乙类传染病的发病率有很强的周期性和季节性,ARIMA模型能很好的描述这些特性,因此针对传染病的发病率数据建立乘积季节ARIMA模型。利用该模型进行预测,结果显示预测效果良好,ARIMA模型可以有效的应用于云南省乙类传染病发病率的预测。第三,针对云南省乙类传染病发病率数据的非线性特性,构建BP神经网络预测模型。BP神经网络预测模型可以很好的拟合传染病发病率的非线性特征,BP神经神经网络具有良好的非线性预测能力,可以用在云南省乙类传染病的发病率预测上。最后,ARIMA模型的残差是白噪声序列,具有高度的非线性特征,而BP神经网络又可以很好的描述这些特征,因此将ARIMA模型和BP神经网络模型进行组合,构成了新的组合模型,ARIMA_BP组合模型。通过预测评价指标MAPE、RMSE和MAE对几种模型进行综合评价,结果表明,ARIMA_BP组合模型在云南省乙类传染病的预测上要优于单一的ARIMA模型和BP神经网络模型。更多还原

【Abstract】 Infectious diseases can do great harm to people’s life.Light can cause discomfort in the patient’s body,and the weight can seriously affect the dysfunction of various organs and even life.Therefore,it is very important and urgent to predict the incidence of infectious diseases more accurately.The prediction of the incidence of infectious diseases can not only provide references for relevant departments,make correct decisions in advance,but also enhance people’s awareness of the risk of infectious diseases,so that infectious diseases can be prevented and controlled to some extent.This paper uses the method of time series prediction to analyze and study the incidence of infectious diseases of class B in Yunnan province.The incidence rate is forecasted by BP neural network and ARIMA model.Finally,a new combination model is formed by combining BP neural network model with ARIMA model.The main work of this paper is as follows:First of all,the grey prediction model was established based on annual population data of Yunnan province.The annual population data of Yunnan province in 2017 and 2018 were predicted with 2005-2016 year data.The results show that the prediction level of the model is outstanding,which shows that the model can be used well in the population data prediction of Yunnan province.Secondly,the incidence of infectious diseases of class B in Yunnan province is very cyclical and seasonal,and the ARIMA model can describe these characteristics well.The ARIMA model is established for the incidence of infectious diseases data.The prediction result is very good by using the model Using this model.The ARIMA model can be effectively applied to the prediction of the incidence of infectious diseases of class B in Yunnan province.Thirdly,in view of the non-linear characteristics of the incidence of infectious diseases of Class B in Yunnan Province,the BP neural network model is established.The BP neural network model has good non-linearity and adaptability,the results show that BP neural network has good predictive ability,The BP neural can be effectively applied to the prediction of the incidence of infectious diseases of class B in Yunnan province.Finally,the residual ARIMA model is white noise sequence,has highly nonlinear characteristics.In order to further improve the accuracy and prediction effect,we combine ARIMA model and BP neural network model to form a new combination model— ARIMA_BP combination model.The several models were evaluated by MAPE,RMSE and MAE.The results showed that the ARIMA_BP combination model was better than the ARIMA model and the BP neural network model in the prediction of class B infectious diseases in Yunnan province.更多还原