[1Yin-A] Poster 1

[1Yin-A-01] Demand Forecasting and Explainability through Similar Product Trend Information

〇Kudo Fumiya¹, Horiwaki Kazuki¹, Kazuo Muto¹ (1. Hitachi)

Keywords:

demand forecasting,rationale generation,large language model

Demand forecasting for product shipments is critical in the manufacturing industry to optimize inventory turnover and reduce excess stock. However, when deploying machine learning-based forecasting in practice, a major challenge lies in enhancing the interpretability and acceptance of predictions among business stakeholders. This study proposes a novel forecasting approach that integrates the Gompertz model with information from products exhibiting similar shipment trends to the target product, along with a rationale generation method leveraging large language models (LLMs). Experimental results indicate that incorporating similar shipments trends improves prediction accuracy and enables the generation of interpretable rationales, such as shipment trend data of comparable products.