Bayesian Intermittent Demand Forecasting for Large Inventories

By Matthias W. Seeger, David Salinas, Valentin Flunkert. 2016

103 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods, Time-Series & Forecasting

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Probabilistic & Bayesian Methods

In Time-Series & Forecasting

Citations in the corpus

DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks (2017, methodology): “For the parts and ec / ec-sub datasets we compare to the negative-binomial autoregressive method of Snyder et al. [19] ( Snyder ) and the ISSM method [18], which are the strongest published re-sults on these datasets.”
Deep Factors for Forecasting (2019, methodology): “The second local model, DF-LDS, is a part of a special and robust family of SSMs, Innovation State-Space Models (ISSMs) (Hyndman et al., 2008; Seeger et al., 2016).”
Think Globally, Act Locally: A Deep Neural Network Approach to High-Dimensional Time Series Forecasting (2019, background): “1 Introduction Time-series forecasting is an important problem with many industrial applications like retail demand forecasting [21], financial predictions [15], predicting traffic or weather patterns [5].”
A Disentangled Recognition and Nonlinear Dynamics Model for Unsupervised Learning (2017, methodology): “LGSSM with dynamic parameterization have been used for large-scale demand forecasting in [27].”
State Space Gaussian Processes with Non-Gaussian Likelihood (2018, methodology): “Similar primitives have been used in Kalman-based demand forecasting (Seeger et al., 2016) with linear models.”
GluonTS: Probabilistic Time Series Models in Python (2019, background): “State Space Models (SSMs) provide a principled framework for modeling and learning time series patterns (Hyndman et al., 2008; Durbin and Koopman, 2012; Seeger et al., 2016).”
Approximate Bayesian Inference in Linear State Space Models for Intermittent Demand Forecasting at Scale (2017, methodology): “terns occur frequently in real data. For example, overall sales may be larger on Sundays than on Mondays, but this ratio may change across the year. While such seasonality patterns are represented in [16] via features in the generalized linear model, they cannot drift, and only the average behavior ”
Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting (2020, background): “Classical time-series models serve as a reliable workhorse for time-series forecasting (Box et al. 2015; Ray 1990; Seeger et al. 2017; Seeger, Salinas, and Flunkert 2016), and deep learning techniques mainly develop an encoder-decoder prediction paradigm by using RNN and their variants (Hochreiter…”
Auto-Differentiating Linear Algebra (2017, background): “Even time series models with non-Gaussian likelihood, such as [20], can be trained in MXNet.”
Forecasting Big Time Series: Old and New (2018, background): “In particular, we focus on the exponential smoothing models (Innovation state space models (ISSM)) [11] and structural time series models [8, 5, 30] along with their Bayesian counterparts [32, 31].”
Deep Factors with Gaussian Processes for Forecasting (2018, background): “This shift can be attributed to the availability of large and diverse time series datasets in a wide variety of fields [2].”
Probabilistic Demand Forecasting at Scale (2017, methodology): “While the parallel execution of this scheme is trivial, it allows us to easily scale out powerful algorithms not covered in this writeup, such as maximum likelihood parameter learning in state space models with non-Gaussian likelihood, using approximate Bayesian inference [26].”
Deep Learning for Time Series Forecasting: Tutorial and Literature Survey (2020, background): “, by [160], have not yet been carried over to NNs.”
Sample Path Generation for Probabilistic Demand Forecasting (2018, background): “[38] also propose a likelihood based latent state model - while the generative adversarial network approach advocated by our work is does not rely on any likelihood assumptions.”
A Unified Framework for Long Range and Cold Start Forecasting of Seasonal Profiles in Time Series (2017, background): “For example, [5] addressed the challenge of predicting intermittent demand of products.”