2266. Intermittent demand forecasting with count panel data models
Invited abstract in session WC-6: Predictive Analytics: Forecasting I, stream Analytics, Data Science, and Forecasting.
Wednesday, 13:30-15:00Room: H9
Authors (first author is the speaker)
| 1. | Katja Nieberle
|
Abstract
Intermittent time series frequently occur in domains such as spare parts demand, service logistics, and slow-moving inventory systems, where demand is irregular and often zero. Most forecasting methods for intermittent demand time series use a univariate approach and estimate model parameters at the individual stock-keeping unit (SKU) level. Due to the high frequency of zero demand, it is often impossible to capture underlying trend- and season-parameters or local patterns, which may be visible at higher levels of product hierarchies.
We present a multivariate approach for forecasting groups of intermittent demand time series with similar patterns. We use count panel data distributions, in which both individual parameters and common parameters at the aggregated hierarchy level are estimated simultaneously.
Our approach is tested against commonly used intermittent demand forecasting methods.
Keywords
- Forecasting
- Predictive Analytics
- Time Series Analysis
Status: accepted
Back to the list of papers