【摘要】 传统时间序列方法在预测模型中要求时序数据稳定,但对复杂的非线性系统拟合能力较差,但GDP增长的预测精度不够准确。为了提高GDP增长的预测精度,首先利用机器学习算法Random Forest对影响GDP增长的变量进行重要性排序,选取重要变量,之后运用深度学习中的LSTM神经网络对GDP增长进行预测分析,并将预测结果与传统时序型ARIMA及GARCH模型进行比较。实验结果表明,基于递归神经网络的LSTM模型能较准确地反映我国GDP增长的变化规律。因此,LSTM模型在宏观经济预测中具有较高的应用价值。更多还原

【Abstract】 The traditional time series method requires stable time series data in the forecast model,and has poor ability to fit complex nonlinear systems,and the forecast accuracy of GDP growth is not accurate enough.In order to improve the prediction accuracy of GDP growth rate,the machine learning algorithm Random Forest is used to rank the importance of the variables that affect GDP growth rate.After that,important variables are selected,and the LSTM neural network in deep learning is used to predict and analyze GDP growth rate.The prediction results are compared with the traditional time-series ARIMA,GARCH models,etc.The experimental results show that the LSTM model based on the recurrent neural network can accurately reflect the change law of China’s GDP growth rate.Therefore,the LSTM model has higher application value in macroeconomic forecasting.更多还原