DEMAND PLANNING FOR BUILDING ENGINEERING PRODUCTS – A CASE STUDY WITH TRANSFORMER-BASED NEURAL NETWORKS.

Efficient demand planning holds critical significance for businesses. In this research, the authors investigate the applicability of the Temporal Fusion Transformer, a neural network-based model, to address demand planning challenges. Specifically, they explore the potential benefits of incorporating additional information related to product characteristics and sales channel types. The primary objective of this study is to assess the advantages gained by incorporating these supplementary variables. The dataset utilized in this analysis originates from a company predominantly engaged in the sale of building engineering products. The authors initially focus on static attributes such as product groupings and time-varying attributes such as sales channel variations. This paper’s contribution lies in its comprehensive case study, which applies the Temporal Fusion Transformer model to a real-world demand planning problem of the company, including all its specifications and customizations. [ABSTRACT FROM AUTHOR]