Graph Neural Networks in Intermittent Time-Series Forecasting

DOI

10.2991/978-94-6463-445-7_13How to use a DOI?

Keywords

Intermittent forecasting; fully convolutional neural networks; graph neural networks

Abstract

Forecasting in the presence of intermittence is a challenging task. Several classic statistical approaches such as ARIMA is not directly applicable to such problem due to the stationary assumption ARIMA has. Fortunately, several methods such as Croston and ADIDA were devised to handle forecasting in the middle of intermittence. Generally, methods in forecasting intermittent time-series can be classified into two groups. The first group used inter-demand intervals and incorporated it in the forecasting model, and the second group used aggregation to obtain smoother time-series data. To the best of our knowledge, there is no model that tries to aggregate the effects of both methods. This paper tries to use fully convolutional neural networks (FCNN) with gating mechanisms to combine the methods present in each of the two groups while keeping the model efficient to train. Furthermore, this paper also tries to incorporate state-of-the-art model in graph neural networks in time-series forecasting to extract spatial information between time-series and incorporate it in the forecasting result, as research in graph neural networks for time-series forecasting is rising in popularity to address spatial dependence. Our model outperforms existing models which are prominent in the field of intermittent demand forecasting, and an ablation study is also done to address the effects of each part of the model. Our study shows that combining methods already established in intermittent demand-forecasting with state-of-the-art model in graph neural networks for time-series forecasting achieves better result in terms of metrics.

Copyright

© 2024 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Mikhael Belmiro
AU  - Finny Oktariani
PY  - 2024
DA  - 2024/06/29
TI  - Graph Neural Networks in Intermittent Time-Series Forecasting
BT  - Proceedings of the 2nd International Conference on Neural Networks and Machine Learning 2023 (ICNNML 2023)
PB  - Atlantis Press
SP  - 108
EP  - 117
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6463-445-7_13
DO  - 10.2991/978-94-6463-445-7_13
ID  - Belmiro2024
ER  -