Igor Jakubiak & Patryk Iwanek - Forecasting the Long Tail - Predicting Intermittent Product Sales
[DSS22 Recordings] [Igor Jakubiak & Patryk Iwanek]
We invite you to watch the recording from last year's Data Science Summit:
Topic: Tail Sales Forecasts - Predictions of Infrequent Product Sales Based on a Hierarchy of Seasonal Patterns
Speakers: Igor Jakubiak & Patryk Iwanek
Description: Generating demand forecasts for infrequent products (intermittent demand) is an exceptionally challenging task. Standard approaches based on time series, machine learning, and even the Croston method in many applications return low-quality results. This presentation will present a proprietary method for predicting demand for products in the so-called "long tail" of the sales distribution, using an example implementation at one of the largest automotive spare parts distributors in Europe. The system generates predictions for nearly 10 million time series based on seasonality patterns calculated at the intersection of product and geographic hierarchies.