Discussion of Multivariate Dynamic Modeling for Bayesian Forecasting of Business Revenue COMMENT

This paper, Multivariate Dynamic Modeling for Bayesian Forecasting of Business Revenue, proposes a novel Bayesian approach based on dynamic linear models to share information from different sectors, LSG (Local Store Group), and item category, through the use of auxiliary information (the discount information). The authors demonstrate the feasibility of parallel computing with multiple item categories, making the Bayesian method highly scalable. The proposed method in the paper should have wide applicability in inventory and revenue management. We suggest in this discussion potential areas for further development.

Automated summary

This paper proposes a novel Bayesian approach based on dynamic linear models for multivariate dynamic modeling, which enables information sharing among different sectors, local store groups, and item categories through the use of auxiliary information. The authors demonstrate the feasibility of parallel computing with multiple item categories, making the Bayesian method highly scalable. The proposed method in the paper should have wide applicability in inventory and revenue management.