Ronald Menich, Chief Data Scientist, Predictix, LLC
summary
Claims that replacing legacy time‑series stacks with factorization machines plus heavy feature engineering yields more accurate, extensible retail forecasts; uses AWS elastic scaling, stochastic‑gradient/LBFGS solvers, ensembles that train on data subsets and reconcile parameters. Works on a retailer with ~150,000 SKUs and 8,000 stores forecasting 52 weeks ahead, handles timing and zero filling only when on‑shelf, models up to 4‑way interactions across price/promos/competition, and learns from datasets with 95% zeros to deliver measurable gains (~1% WAPE → ~0.75% inventory ← ~$125M savings).
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:20Predictix Team
- 0:57Decision Support
- 4:26Demand forecasting
- 5:30Time series models
- 8:01Retail environment
- 9:04Competition
- 10:39Machine Learning
- 15:36Timing
- 17:56Data Dynamics
related talks
31mFrom stockouts to happy customers: Proven solutions for time series forecasting in retail
40mMulti-Series Forecasting at Scale with StatsForecast
40mDomain-driven Data Science
35mBuilding a successful inventory planning solution with Amazon Forecast
29mProbabilistic demand forecasting at scale
22mFrontiers of Machine Learning from the Lens of Retail
28mHow Human Expertise and AI drive Scalable Forecasting Success at
39mProductionising Data Science at Scale