不同时间序列分析方法在高血压发病率预测中的比较

Application of time series analysis in the prediction of hypertension incidence

目的比较时间序列分析中指数平滑法、移动平均法、自回归分析及求和自回归滑动平均法(ARIMA)在海西州地区2001年1月至2007年12月高血压发病率预测中的优劣。方法用时间序列分析各方法建模预测,比较各方法的误差平方和、赤池信息量、施瓦茨信息量、拟合优度和拟合效果,确定最佳预测方法。结果指数平滑法、移动平滑法、自回归分析法及求和自回归滑动平均法中,ARIMA模型的误差平方和、赤池信息量、施瓦茨信息量最小,拟合优度最大,拟合效果最好。结论海西州地区2001年1月至2007年12月的高血压发病率预测中,时间序列诸分析方法中求和自回归滑动平均法预测效果最好。

Objective To provide the fittest model for forecasting hypertension incidence in Haixizhou region by comparing the results of exponential smoothing, moving average, autoregressive model and autoregressive integrated moving average model （ARIMA model） from January to December in 2007. Methods The fitting effects of four statistical methods were compared and the fittest model was chosen. Results Among the four statistical methods, the effect of fitting and prediction of moving average model was the best. Conclusions Moving average model seems to be the fittest one in the prediction of hypertension in Haixizhou region from January, 2001 to December,2007.