ENERGENIUS
Published October 15, 2025 | Version v1
Journal article Open

Forecasting intermittent time series with Gaussian Processes and Tweedie likelihood

Authors/Creators

  • 1. ROR icon Dalle Molle Institute for Artificial Intelligence Research

Description

Peer-reviewed journal article published in the International Journal of Forecasting proposing Gaussian Process models with negative binomial and Tweedie likelihoods for probabilistic forecasting of intermittent time series. The approach improves high-quantile estimation and uncertainty modelling for supply chain demand forecasting.

Files

1-s2.0-S0169207025000974-main.pdf

Files (7.9 MB)

Name Size Download all
md5:d7aa52d61cc360543adca73b9ecde9e7
7.9 MB Preview Download

Additional details

Related works

Is identical to
Journal article: 10.1016/j.ijforecast.2025.10.001 (DOI)

Funding

European Commission
ENERGENIUS - leveraging the energy transition by gamified learning and AI, guided by cross sectoral integrated services and digital twin models to foster accessible and human-centered energy saving experiences 101160720
Swiss National Science Foundation
200021_212164/1