Forecasting Chinese GDP with Mixed Frequency Data Set: A Generalized Lasso Granger Method

In this paper, we introduce an effective machine learning method which can capture the temporal causal structures between irregular time series to forecast China GDP growth rate with Mixed Frequency data set. The introduced method first generalized the inner product operator via kernels so that regression-based temporal casual models can be applicable to irregular time series, then the temporal casual relationships among the irregular time series are studied by Generalized Lasso Granger (GLG) graphical models. The main advantage of this approach is that it does not directly estimate the values of missing data of low frequency time series or has restricted assumptions about the generation process of the time series. By applying this method to a 17 macroeconomic indicators GLG model, the forecasting accuracy is better than the autoregressive (AR) benchmark model and a widely used mixed-data sampling (MIDAS) model.