基于季节调整的BP神经网络等三种模型对中国出口研究

本文以出口额、实际汇率、我国GDP、美国IPI及它们的季节变量等六个变量为决定变量,运用BP神经网络、ARIMA及AR-GARCH三种方法,对我国向美国的出口额分别建模,并进行了预测。选取误差指标,分别对三个模型得到的模拟结果和预测结果同真实值进行比较。结果发现,三种模型效果都令人满意,虽在模拟和预测能力上有一定差别,但ARIMA模型优势明显。本文分析了以上结果产生的原因,并结合模型为提高我国出口提出建议。

Study on China Exports through BP Neural Network, ARIMA and AR-GARCH Based on Seasonality

By using the first order lag of export value, the first order lag of real exchange rate, the synchronization of last year of real exchange rate, etc, as determine variables, this paper sets up three models, BP neural network, ARIMA and AR-GARCH, to simulate and forecast exports from China to America. Then it adopts the criteria MAPE to compare the performance of these models. The results reveal that all three models perform well while different model have different characteristics, and ARIMA shows the best ...