Quarterly GDP Forecasting Application Based on Graph Attention LSTM Deep Learning
GDP forecasting has always been an important research topic in the field of macroeconomics.At the same time,integrating deep learning algorithms and monitoring the dynamic changes of GDP in real time has been an inevitable trend in macroeconomic indicator forecasting.Based on this,this paper proposes a new GAT-LSTM fusion deep learning model and forecasts quarterly GDP considering the multi-source nonlinear spatio-temporal characteristics of macroeconomic variables.The model uses a graph attention network(GAT)to capture spatial topological structure information and a long short-term memory neural network(LSTM)to extract time series information to improve the model prediction effect.The results show that compared with the rest of the benchmark models,the GAT-LSTM model proposed in this paper has good generalization ability and robustness,with an average improvement of 0.592 9 in regression fit(R2)and a decrease of 0.6617 in mean square error(MSE).The model has a good application prospect in the field of GDP forecasting for investors,enterprises,and the state to make scientific decisions and improve economic efficiency.
GDP forecastinggraph attention networklong short-term memory neural networkdeep learning
万方数据
维普
