RoME for complex hierarchies

Background: The RoME methodology is primarily developed and demonstrated for standard hierarchical time series (HTS) defined by bottom-level aggregation constraints. However, time series applications often involve more intricate structural relationships.

Question / Future Work: Extending the proposed RoME framework to accommodate more complex hierarchical structures, such as temporal and cross-temporal/sectional hierarchies, represents a significant direction for future research. Furthermore, explicitly incorporating non-negativity constraints, which are common in certain forecasting domains, into the reconciliation process also merits investigation.