Study On Epidemic Trend And Time Series Model Of Prediction For Human Brucellosis In China

Objective:Study on trend analysis of three dimension distributions in the inc idence of human brucellosis from 2004 to 2012 years, to determine high-risk time, place and people. Menwhile,to research ARIMA model, BPNN model and ARI MA- BPNN hybrid model in the application of the outbreak early warning.Finally, the appropriate model was selected from the three models was used forecasting human brucellosis in China, provided a scientif ic basis for prevention and control work.Methods: Studyed trend analysis of three dimension distributions for human brucellosis in China,the some method were used to do that X-12-ARIMA seasonal adjustment, Joinpoint regression anlysis and the dynamic sequence,to analysis of seasonal epidemic, the trends and short-term volatility. ARIMA model, BP neural network model and ARI MA- BPNN combination model were used to predicte outbreaks.Then, according to the forecast precision index, selected the optimal model, predicted brucellosis epidemic situation form January to June in 2006 years,in China.Results:From 2004 to 2012 human brucellosis trend results show: the number and incidence of disease, the highest and lowest were May and December every year, overall were on the rise;More men than women in a year, onset age focused on between 35 and 45 years old;professional was given priority to with farmers and herdsmen;regional distribution, the top five provinces each year respectively were Inner Mongolia, Hei longjiang,Shanxi and Hebei and Jilin province.In the forecast of epidemic to forecast Numbers for the onset of object, respectively established ARIMA model, BP neural network model and BPNN- ARIMA model, through comparing the accuracy of prediction index. Finally, selected optimal combination model of ARI MA- BPNN model, the MAPE=4.84, R2 =0.9862, forecasted January to June in 2016 monthly toll of disease: 3311, 3068, 5746, 6477, 7303 and 8196.Conclusion:In China,human brucellosis of outbreaks still rise in recent ten years, incidence of age changed more small.Among people of professional, the priority were farmers and herdsmen,and regional distribution were more extensive. For ARIMA- BPNN hybrid model, prediction accuracy index were superior to the single ARIMA model and BP neural network model, it may predicte the number of human brucellosis, and be used as a kind of method to warning for human brucellosis, in China.