论文部分内容阅读
目的:立足卫生支出等趋势预测问题时间序列组合建模研究,借助算例比较验证。方法:根据政府卫生支出时序资料,将曲线拟合法和ARIMA法建模以后,考虑了残差修正组合法和线性加权组合法(残差平方和倒数法、关联度法、相关系数法、待定系数法和等权法)两大类组合途径,共建立六个组合模型。结果:修正指数曲线、ARIMA法拟合及预测效果不错;残差修正组合法实际算例不必要,其余五种线性组合模型效果优于任何单个方法模型。结论:曲线法、ARIMA法及六种组合模型对于趋势预测有应用意义。
OBJECTIVE: To study the time series combination modeling of trend forecasting problems such as health expenditure, and to compare with the example. Methods: According to the government health expenditure time series data, after the curve fitting method and the ARIMA method were modeled, the residual error correction combination method and the linear weighted combination method (residual sum of squared reciprocal method, correlation degree method, correlation coefficient method, undetermined coefficient Law and Equal Rights Law) two kinds of portfolio approach, a total of six portfolio model. Results: The modified exponential curve and ARIMA method fit well and the prediction result is good. The actual example of residual error correction combination method is unnecessary, and the other five linear combination models are better than any single method model. Conclusion: The curve method, ARIMA method and six kinds of combined models have potential significance for the trend prediction.