Applying Time Series Model To Fit And Predict HFRS & AIDS Trend

There axe certain regulation in the emergence of disease, so the predict technology can discover the disease development trend early. That can remind the epidemiology expert and the staff member to adopt the countermeasure of controlling in time. The mathematics model of epidemiology is on the basis of the prevailing course of some disease , the main factors and the interactions on influence, using the prevail characteristic of course and mathematics expression formula ration to explain and reflect disease ecological mathematics relational expression that restriction concerns of quantity, it is the mathematics simulation of the disease propagation course. The mathematics model of epidemiology is applied to each field of epidemiology studies extensively, such as studying the appraisal of the prevailing characteristic, result of the disease and prediction of the disease. The model has important function.This research tried to regard mathematics model of epidemiology as the basic method. The materials based on the infectious disease epidemic situation of Shenyang of 1984 - 2004. We described quantitatively the prevailing laws and the development of the main diseases in Shenyang city, and probed into applying mathematics model to fit the actual distribution of the diseases, and offered the theoretical foundation for making the science, effective control and prevention measures.Materials and Methods1. Materials: This materials stem from infectious disease materials andpopulations materials from the 1984 to 2004 in Shenyang centre for disease control and prevention.2. Methods; (1) method principle; Steady time series analysis. The method to set up compound model is using least square method fit determinacy part of data array, begin from the low steps , increase the steps to count gradually, until the model has not been obviously improved . Then set up suitable autoregres-sion - rolling moving average model in dispelling the incomplete quantity array which confirms the trend ( ARM A (p, q) ). Finally, use above - mentioned two parts of parameters valuation as initial value, estimate parameters of determinacy part and of ARMA part again with non - linear and two minimum methods, draw the final estimation of compound model.(2) We use various kinds of compound models; narrate as follows respectively from simple situation to general situation: 1 ) Linear trend. 2) Index trend. 3) Cycle trend. ARMA is only suitable for describing steady time arrays, but the time array met in practical application is often non - steady. When the random array is not approximate sometimes, carry on difference and divide operation, it can turn the non - steady array into the steady array or approximate steady array, the melting steadily of some arrays needs numerous difference to divide dealing with too.3. Analysis tools: DPS (Data Processing System) from the Zhejiang University.Results1. The fit and prevail trend prediction of HFRS(1) Model discerns: Drawing the time data distribution map of the array according to an annual HFRS incidence of disease of year 1984 - 2004 (1/ 100000) , carrying on straight attitude inspection to the data. The initial data array need to carry on the data changing, and normal distribution appears.(2)Model steps distinguish and diagnose; Choose the principle according to model steps of ARMA, select ARMA (1,0), ARMA (1,1) and ARMA (0, 1) for use separately. The model of ARMA (1,1) is ideal according to theStandard of the judging standards.(3) The comparison of predicted values and actual value: The absolute value of the absolute errors of predicted value and actual value is 0. 037 per 100000, 0. 955 per 100000, 1. 477 per 100000 respectively. It is 0. 823 per 100000 on average.(4) Prediction of incidence of disease ?. the predict result of disease incidence in 2005 is 8. 85 per 100000.2. The fit and prevail trend prediction of AIDS(1) Model discerns; Drawing the time data distribution map of the array according to an annual AIDS amount of disease of year 1991 -2004, carrying on straight attitude inspection to the data. The initial data array does not become normal distribution, so need to carry on the data changing, normal distribution appears.(2) Model steps distinguish and diagnose; Choose the principle according to model steps of ARIMA, select ARIMA (1,1,0), ARIMA (1,1,1) and ARIMA (0,1,1) for use separately. The model of ARIMA (1,1,1) is ideal according to the Standard of the judging standards.(3) The comparison of predicted values and actual value: The absolute value of the absolute errors of predicted value and actual value is 0. 00686 per 100000.(4)Prediction of incidence of disease; 2005 -2007 predict value respectively is: 0.73162 per 100000, 1.03194 per 100000, 1.42632 per 100000.Discussions1. The time array prediction model supposes the target's change only relates to time, according to its changing characteristics, infer its future state with the inertia principle. By contrast with Box -Jenkins model, ARIMA model is simple, it is effectual to predict, and especially a lot of statistics software can be used for ARIMA procedure, suitable for personnel in non - mathematics field especially.2. Any Prediction method can have error; through infectious disease pre-diet to train the signal, according to predicting result and deviation between the situations actually making the judgment. In the world people appraise and predict the result through calculating sensitivity, peculiar degree, positive predicted value, negative predicted value and coincidence rate too at present.3. Prediction is served for preventing and controlling the disease, if prediction, is for controlling the breaks our and prevailing of recent disease, it can be confirmed as short - tern/ forecast ,if prediction is for making the prevailing control strategy of disease, it can be confirmed as long - term forecast . So, this research has carned on short - term lorecast to HFRS, this keeps the same with document report.4. Because historical materials used stemming from the legal infectious disease and report the system and cause of the death and report system mainly, and time span report system go through several change. We should consider integral-ity and credibility fully, explaining and using its result with discretion. The model needs io be revised with the new data constantly, dispel the influence of the unusual data gradually.5, This text carries on the prediction of disease incidence in " year" on HFRS, the precision is relatively gold, it often need predict by moon in the real work, with and materials quality improvement, the disease prediction to go on by week and moon must be developing direction,6. This research adopt experience law while choosing and predicting one a-head of time, it have not screened to one ahead of time , propose improving in the following work.7, For the purpose oir achieving the early warning better, it should not merely predict the disease; it should certainly predict the relevant factors of the disease in the future ARIMA. is a single independent variable model, it does not consider the complicated relation between influence factor connection and restriction , it seem powerless to the disease that popular factor is complicated.ConclusionsARIMA model has not required strictly to capacity of the sample and proba-...