Earticle

초록

영어

Despite the complexities of forecasting demand in the retail sector, prediction models need to account for essential temporal and operational variables to ensure robust scalability. Nevertheless, identifying multi-level temporal patterns across retail channels and deciphering the interplay between static industry attributes and dynamic sales trends remain challenging. To tackle these problems, we propose a multiview learning framework specifically designed for distribution channels. By incorporating both local and global patterns, our framework captures a holistic view of retail characteristics and potential sales trends. We validate the effectiveness of our method using real-world data from a beverage distributor, showcasing its superiority compared to conventional techniques. Furthermore, we highlight the critical role of understanding the interaction between local and global patterns. This research presents a decoupled forecasting framework that enhances accuracy and provides valuable insights for academic researchers and actionable guidance for retail practitioners.

목차

Abstract
Introduction
Literature Review
Demand Forecasting for Distribution Channels
Methods
Supply and Operation Planning Process
Data and Features
Multiview Learning Framework
Predictive Analytics and Evaluations
Preliminary Analyses
Future Research Direction
References

저자

• Seongbeom Kim [ Graduate School of Information, Yonsei University ]
• Jaehui Kim [ Graduate School of Information, Yonsei University ]
• Hee-Woong Kim [ Graduate School of Information, Yonsei University ]

참고문헌