ARMA模型与bilinear模型的比较

Comparison of ARMA Model and Bilinear Model

文章从参数估计和数据拟合的视角出发，对时间序列中两个经典的模型-ARMA模型和bilinear模型的特性进行比较，比较之后发现：对于线性结构的数据，ARMA模型与bilinear模型拟合准确度相近，而对于具有非线性结构的数据，bilinear模型的拟合准确度明显高于ARMA模型。而且随着非线性部分在数据中占比的提高，bilinear模型相对于ARMA模型拟合的优越程度有所增加。

Starting from the perspective of parameter estimation and data fitting, this paper makes a comparison on the two classical models in time series： ARMA model and bilinear model. The result shows that ARMA model and bilinear model have similar fitness on the data for linear structures, but bilinear model has a significantly higher fitness than ARMA model to the data for nonlinear structures, and that with the increase of the proportion of nonlinear parts in the data, bilinear model has an increased superiority to ARMA model to an extent.