基于AR-ANN模型的指数预测及脉冲响应函数预警

Forecast Based on AR-ANN Model and Warning from Impulse Response Function

融合时间序列与神经网络算法,得到AR-ANN模型。结合多维时间序列在捕捉线性关系的优势,以及神经网络在预测非线性关系的优势,AR-ANN模型对上证指数有较好的预测效果。另建立结构向量自回归模型(SVAR),反应了指数受其他经济内生变量的一个随机冲击后的变化状况,通过脉冲响应函数对预测做出预警。

Multiple economical endogenous variables exist in relation to the Shanghai composite index which is non-linear and tends to be erratic. This paper established the AR-ANN model which combined time series model and artificial neutral network. Owing to the strength of linear forecast of AR model and the non-linear forecast of ANN, the forecast made by the AR-ANN model proved satisfactory. Further, it established the Structural Vector Auto Regression model in an attempt to identify the specific mechanism in which Shanghai composite index responds to the random shock from other economical endogenous variables. This model generated forecast of higher accuracy and issues warning from impulse response function.