EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

NeuralProphet for Ordinal Time Series Forecasting: A Two-Stage Approach with Threshold Optimization

Yuna Han, Amira Abdul-Rahim, Sinclair Nzenwata, Drew Lazar

Forecasting · 07.11.2026 · 8 (6) · 105

DOI: 10.3390/forecast8060105 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Forecasting, online erschienen am 07.11.2026. Verfasst von Yuna Han und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Air quality indices, energy load levels, and similar series are recorded as ordered categories, yet standard forecasting methods produce continuous predictions. We propose a two-stage approach using NeuralProphet, which decomposes a series into trend, seasonality, and autoregressive components. In the first stage, NeuralProphet is fit to the ordinal labels treated as continuous, yielding a decomposition and a continuous forecast. In the second, ordered thresholds mapping it to categories are optimized on a validation segment, replacing naive rounding. We evaluate the method on simulations and on two hourly series, Beijing air quality and electricity load in a Ball State University building, with U.S. Drought Monitor categories as a third series. In simulation the method recovers the latent ordering and generating components, and threshold optimization improves the quadratic weighted kappa over rounding in every condition, especially when the first stage is miscalibrated but order-preserving, as under trend shifts or evolving seasonality. On the hourly series it outperforms logistic and ordered-logit baselines at every horizon, and threshold optimization improves on rounding at all but the shortest Beijing horizon and at every drought horizon. The approach is an effective, interpretable route to ordinal forecasting without a specialized ordinal model.

Zitieren

APA

Han, Y., Abdul-Rahim, A., Nzenwata, S., & Lazar, D. (2026). NeuralProphet for Ordinal Time Series Forecasting: A Two-Stage Approach with Threshold Optimization. Forecasting, 8(6), 105. https://doi.org/10.3390/forecast8060105

Vancouver

Han Y, Abdul-Rahim A, Nzenwata S, Lazar D. NeuralProphet for Ordinal Time Series Forecasting: A Two-Stage Approach with Threshold Optimization. Forecasting. 2026;8(6):105. doi:10.3390/forecast8060105

BibTeX
@article{Han2026,
  author = {Han, Yuna and Abdul-Rahim, Amira and Nzenwata, Sinclair and Lazar, Drew},
  title = {NeuralProphet for Ordinal Time Series Forecasting: A Two-Stage Approach with Threshold Optimization},
  journal = {Forecasting},
  year = {2026},
  volume = {8},
  number = {6},
  pages = {105},
  doi = {10.3390/forecast8060105}
}
Kostenlos starten →