Machine Learning with a Deep Dive into Automated Forecasting System
summary
Walmart builds an automated self-learning forecasting system that runs multiple models per SKU per store at scale (roughly 5,000 US stores and millions of items, 260M+ transactions weekly producing petabytes). It ingests internal and external features via Python web crawlers, stores/processes in Hadoop, Hive and R, fits ARIMA/DLM/Fourier and home‑grown models, then store-item level forecasts are ensembled with weights that update from back‑testing. Metrics include MAPE and hit/miss counts over 3–12 week horizons; the pipeline stabilizes and improves forecasts while aiming to reduce buffer inventory.
AI-written summary · about
chapters (12)
- 0:00Machine learning at Walmart
- 2:34Machine learning applications across Walmart's ecosystem
- 5:40Traditional ML vs forecasting
- 8:28Need for automated forecasting
- 10:53Self-learning forecasting system
- 15:26System architecture & learning
- 16:03Q&A: Customer mapping & data
- 16:48Q&A: Forecasting validation
- 17:36Q&A: Supply chain integration
- 18:18Q&A: New product forecasting
- 19:47Q&A: Variable change impact
- 21:11Q&A: Model selection
related talks
09mDemand Forecasting in Supply Chain Using AI Basis Machine Learning
35mBuilding a successful inventory planning solution with Amazon Forecast
28mHow Human Expertise and AI drive Scalable Forecasting Success at
39mProductionising Data Science at Scale
08mMachine Learning for Bangladeshi Retailers
28mMastering Demand Forecasting: Lessons from Europe's Largest Retailer
19mOperational Challenges in Machine Learning Model Life Cycle
16mTransforming Inventory Forecasting with Production ML