Will Machine Learning Improve Forecasts of Economic Growth in China during COVID Pandemic? Gradient Boosting and Random Forest Approaches

This study applies two machine learning models, gradient boosting and random forest, on forecasting annual real GDP growth in China during the COVID pandemic. The forecasts by the International Monetary Fund are used as benchmarks to examine whether machine learning models can improve accuracy of prediction during this period. For each year from 2017 to 2021, the optimal forecasting model is constructed by cross validation and hyperparameter tuning. The accuracy of forecasts is measured by mean absolute percentage error and root squared mean error. The results of this study shows that forecasts by both gradient boosting and random forest are more accurate than IMF benchmarks before COVID pandemic (2017 and 2018), whereas only gradient boosting continues to outperform IMF benchmarks during COVID pandemic from 2019 to 2021. This study provides a valuable reference for the selection of machine learning models on forecasting economic growth, especially under the structural break caused by economic shocks.