HIV/AIDS Infection In Distribution In Hunan Province And To Build Up The Combination Prodiction Model On New HIV Infections

Objective: understanding the basic situation of HIV/AIDS infection in Hunan province,to explore the distribution and whether there have spatial correlation by studying the spatial distribution of HIV/AIDS infection situation.On the basis of time series analysis,grey models and neural network model to set up prediction model in HIV new infectors,we established combined models to improve the fitting precision of the model.Explore its applicablity in Hunan province for HIV new infectors.Methods: the data of HIV/AIDS infection from 2005 to 2015 of Hunan province come from The Center for Disease Control and Prevention(CDC).Population,economy,and social factors obtained through Statistical Yearbook of Hunan province.We using Arc GIS to analysis spatial distribution,to get the intuitive distribution figure;using Open Geoda to analysis the spatial correlation analysis,using moran’I to judge the spatial correlation of infection.Using GDP per capita,per capita medical institutions and beds,employment rate,per capita practicing doctors and city level to make spatial regression analysis,to study the influence on new HIV infection,and try to use Open Geoda to set up spatial panel data model.The study used infection as the training sample data in 2005-2014,2015 data forecast model is set up as a test sample.Time series model is set up by SPSS 19.0,grey model and neural network model is set up in Matlab,using residual error correction,simple combination,simple average combination,weighted average combination to set up combination model.Results: the spatial analysis found that the cumulative HIV infection rate of Hunan province in Hengyang,Yongzhou,Zhangjiajie is the highest,and which is followed by Changsha and Huaihua.Zhuzhou,Xiangtan,and Yueyang infection rate is lowest.AIDS infection rate in Hengyang and Zhangjiajie is higher.Xiangxi,Zhangjiajie and Huaihua which AIDS infection rates are higher.Spatial regression analysis using GDP per capita,per capita medical institutions and beds,per capita medical practitioners(assistant),city level and employment rate on the impact of HIV infection,found that the per capita medical institutions,the per capita bed have correlation with disease.Using Arc GIS to analysis the spatial correlation for the province’s 14 city,HIV,AIDS and HIV/AIDS infection status,the result shows in 2013 and 2014 AIDS infection is space aggregation,Moran’I were 0.2618 and 0.2610,P < 0.05,have statistically difference,indicating the space are related.Using the six factors to analysis the relativity of HIV infection in Hunan province,the result show there no spatial correlation of its(LM lag and LM error,P values > 0.05),so we can’t use spatial panel data to establish spatial lag model and spatial error model.Time series model shows the prediction data within the confidence interval in 2005-2014 HIV infections,the prediction data is slightly different to real value of HIV infection data in Hunan province in 2015,but in the 95% confidence interval,the relative error is low.Establishing grey model,and found that the grey model has the trend of exponential increase,can better response the number of new HIV infections of Hunan province,but months has certain fluctuation trend cannot fit better.Neural network model can fitting the infections,neural network model is better than the grey model fitting effect.The neural network model 2-6-1can in line with the change tendency of the actual data of HIV infections in Hunan province.Set up combination model found that combination model can optimize model to a certain degree,and improve the fitting.Time series model combined with neural network model result show: BP>BPARIMA>WBPARIMA>ARIMAREC>ARIMA+BP>ARIMA;Grey model combined with neural network model show: BP>GMREC>WBPGM>BPGM>GM+BP>GM.Conclusion: the HIV/AIDS infection in Hunan province has certain space aggregation,the per capita number of medical institutions and bed has correlation to disease infection.To predict the number of new HIV infections,get the time series model,grey model and neural network model can better fit the HIV infection situation of Hunan province,the neural network model had best fitting effect.Using combination model can get better fitting effect.