基于Gevers-Wouters算法的ARMA模型改进

Improvement of ARMA Model Based on Gevers-Wouters Algorithm

文章针对在时间序列分析中构造自回归滑动平均模型（即ARMA模型）时存在预测结果不理想的问题,引入Gevers-Wouters算法对模型进行改进。以我国国内航线的旅客周转量的数据为例,剔除数据中存在的季节波动后用BIC准则对ARMA模型定阶,利用Gevers-Wouters算法对模型中MA部分的参数进行调整,由此为之后每一个时期构造各自独有的ARMA模型。通过对模型改进前后预测值以及真实值的比较,预测的准确性有了明显提高。

Concerned on the problem that the prediction results are not ideal when the autoregressive moving-average model （ARMA model） is constructed in time series analysis, this paper introduces Gevers-Wouters algorithm to improve the model. Tak- ing the data of Chinese domestic airline passenger turnover as an example, the paper uses BIC criterion to realize orders determi- nation of ARMA model after eliminating the seasonal fluctuations in the data, and then employs Gevers-Wouters algorithm to ad- just the parameters of MA part in the model and construct unique ARMA model for every period therefrom. By comparing the pre- dictive value with the real value before and after improving the model, the accuracy of prediction is increased obviously.