Prediction On Monthly Prevalence Of AIDS In China Using Intelligent Combination Model

Due to the impact of AIDS epidemic on society,accurate AIDS prediction is of great significance to China.In this paper,two sequences of HIV and AIDS were studied respectively,the machine learning method based on sequence decomposition-recombination was used for prediction.Firstly,original data of HIV incidence in each month can be decomposed through Sin-gular Spectrum Analysis(SSA),obtaining the two parts:reconstruction sequence and resid-ual sequence.Support Vector Regression(SVR)whose parameters c and g are optimized by PSOGSA is used for forecasting reconstruction series,and S-ARIMA is used for forecast-ing residual series.The forementioned method is called combination model SSA-PGSVR-SARIMA.In addition,Support Vector Regression(SVR)whose parameters c and g are op-timized by PSOGSA is used for forecasting reconstruction series,and Extreme Learning Ma-chine(ELM)is used for forecasting residual series,combined them as combination model SSA-PGSVR-ELM.These two combination models can predict the incidence of HIV well.Secondly,a combination model CEEMD-PSR-BPNN is proposed for predicting the month-ly incidence of AIDS,in which Complete Ensemble Empirical Mode Decomposition(CEEMD)algorithm is used to decompose the sequence into five IMFs and one residual series.Then,the Phase Space Reconstruction(PSR)based on C-C method is used to reconstruct the six se-quences separately.Finally,the six groups of reconstructed parts are input into BPNN,and the final AIDS prediction results are obtained.By studying the HIV and AIDS sequence separately and exploring more suitable forecast-ing methods for AIDS,the combination model can effectively predict the prevalence of AIDS and cover the shortage of the previous statistical prediction of infectious disease.The research can provide relevant data support for AIDS prevention and control in China,and the research results have good application value.