Hey everyone,
I'm a final-year B.Tech CSE (AI/ML) student working on a multi-agent supply chain optimization project. I'm currently focusing on the Demand Forecasting Agent and would appreciate some advice on model selection and training strategy.
We're working with a synthetic retail supply chain dataset containing approximately 1.1 million records across 40 different locations .
The goal is to predict product-level demand for each location as accurately as possible, so the forecasts can later be used by other agents for inventory planning and supply chain decisions.
I have a few questions:
Global vs local models: Since the dataset contains 40 locations, should I train 40 separate models , one for each location, or a single global model trained on data from all 40 locations simultaneously , using location identifiers and location-specific features?
Location-specific patterns: If I use a global model, how effectively can it learn differences in demand patterns between locations?…