基于自回归XGBoost时序模型的GDP预测实证

GDP是反映一个国家国民收入、居民消费能力和经济发展的重要宏观经济指标,也是制定相关经济政策的重要依据.选择合适的统计方法研究GDP的发展变化规律,进行短期的高精度预测,对我国的宏观经济决策具有重要意义.研究选用基于自回归的XGBoost时序模型对我国1978-2018年GDP进行拟合预测,Rstudio软件运行结果显示,XGBoost时序模型比经典的时间序列预测模型ARIMA模型、BP神经网络模型、贝叶斯时序模型具有更高的预测精度.在此基础上,运用XGBoost时序模型对我国2019-2023年的GDP进行短期预测,研究结果显示,未来5年我国GDP依然保持持续稳定增长趋势.

An Empirical Study of GDP Forecast Based on the Auto-regressive XGBoost Time Series Model

GDP is an important macroeconomic indicator reflecting a country’s national income,residents’ consumption capacity and economic development,and also an important basis for the formulation of relevant economic policies.It is of great significance for China’s macroeconomic policy-making to select appropriate statistical methods to study the law of GDP development and change and to make short-term high-precision prediction.Based on the historical data analysis of China’s GDP from 1978 to 2018,this paper first chooses autoregressive XGBoost time series model to fit and forecast.The empirical results show that this model has higher prediction accuracy than ARIMA model、BP neural network model and Bayesian time series model.On this basis,the XGBoost time series model is used to forecast the GDP of China in 2019-2023.The research results show that China’s GDP will continue to grow steadily in the next five years.