Bayesian forecast combination using time-varying features

In this work, we propose a novel framework for density forecast combination by constructing time-varying weights based on features. Our estimates in the via Bayesian log predictive scores, which optimal is determined time series features from historical information. particular, use an automatic variable selection method to identify importance of different To end, our approach has better interpretability compared other black-box forecasting schemes. We apply stock market data and M3 competition data. Based structure, simple maximum-a-posteriori scheme outperforms benchmark methods, can further enhance accuracy both point forecasts forecasts.