Abstract：The sudden outbreak of COVID-19 has effected great impact on China's economy, and major economic indicators have declined significantly. However, with the continuous promotion of resumption of work and production, China's major economic indicators show a momentum of recovery and continuous improvement. Monitoring the current state of economic operation and forecasting the future economic trend are undoubtedly of great practical significance for the formulation and timely adjustment of economic policies. This paper constructs a mixed-frequency data regression(MIDAS) model, then estimates 180 models based on different weight function forms and various lag terms of monthly explanatory variables, and finally selects the one with the lowest RMSE as the optimal prediction model. Through calculation, it is concluded that industrial added value has the best prediction effect on GDP. In the sample period, the root-mean-square error is the lowest when is selected the ladder-type weight function,which is only 0.066.

Keyword：MIDAS model; real-time forecast; root-mean-square error;