The slides for my presentation at the International Symposium on Forecasting 2020 are available here.
Building on the work by Pangiotelis et al. (2020) we investigate the implications of the geometric interpretation of hierarchical forecasting further. We propose a new framework for generating hierarchical forecasts, which encompasses previous hierarchical methods while providing insights on their behaviour. A key takeaway is that it is possible to obtain more efficient solutions to the hierarchical forecasting problem. Nonetheless, even though these may be more efficient, the optimisation remains non-trivial. In this work, we identify a series of approximations that balance the efficiency gains with the optimisation complexity to provide superior hierarchical forecasts, as evidenced in our empirical evaluation.
You can find the slides here.