EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Frequency-aligned loss and spectral filtering improve long-range influenza forecasting

Tianyi Feng, Chunyan Luo, Yu Huang

Frontiers in Public Health · 12.01.2026 · 13

DOI: 10.3389/fpubh.2025.1721883 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Public Health, online erschienen am 12.01.2026. Verfasst von Tianyi Feng und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Rehabilitation, West China Hospital Sichuan University Jintang Hospital, Jintang First People's Hospital, Chengdu, China. Hauptthemen laut MEDLINE: Influenza, Human. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Long-horizon forecasts of seasonal influenza remain limited by (i) rapid error growth beyond a few weeks, (ii) entanglement of persistent seasonal cycles with transient outbreaks, and (iii) training objectives that ignore strong autocorrelation in future incidence labels.

Methods

We introduce a frequency-aware pipeline that couples a Spectral Adaptive Filtering Network with a Frequency-Aligned Direct Loss. The backbone first isolates stable global spectral bands and then builds window-specific cross-covariate filters to capture transient events; this convex loss function simultaneously supervises prediction results in both the time domain and the approximated decorrelated frequency domain, effectively reducing bias caused by autocorrelation without sacrificing point accuracy.

Results

On 49 US states (2010–2020), 10 HHS and 9 Census regions (2002–2020), the proposed model lowers MSE by 6–15% and MAE by 2–20% at 24-week horizons vs. six recent baselines while maintaining interpretable band-pass responses that match annual and semi-annual epidemiological periodicities. Ablation and sensitivity analyses confirm that joint time-frequency supervision and dual static-dynamic filtering are both required for peak performance.

Conclusions

Explicit spectral decomposition coupled with autocorrelation-aware training offers a principled route to stable, interpretable long-range influenza forecasting; the modular objective can be plugged into alternative architectures to gain similar error reductions.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Influenza, HumanForecastingSeasonsHumansUnited StatesPrediction AlgorithmsDisease Outbreaks

Stichwörter der Autor:innen

frequency-aligned losslong-horizon epidemic predictionlong-horizon forecastseasonal-influenza forecastingspectral adaptive filtering

Autor:innen und Institutionen

  • Department of Rehabilitation, West China Hospital Sichuan University Jintang Hospital, Jintang First People's Hospital, Chengdu, China.
  • Jilin Institute of Physical Education, Changchun, China.

Zitieren

APA

Feng, T., Luo, C., & Huang, Y. (2026). Frequency-aligned loss and spectral filtering improve long-range influenza forecasting. Frontiers in Public Health, 13. https://doi.org/10.3389/fpubh.2025.1721883

Vancouver

Feng T, Luo C, Huang Y. Frequency-aligned loss and spectral filtering improve long-range influenza forecasting. Frontiers in Public Health. 2026;13. doi:10.3389/fpubh.2025.1721883

BibTeX
@article{Feng2026,
  author = {Feng, Tianyi and Luo, Chunyan and Huang, Yu},
  title = {Frequency-aligned loss and spectral filtering improve long-range influenza forecasting},
  journal = {Frontiers in Public Health},
  year = {2026},
  volume = {13},
  doi = {10.3389/fpubh.2025.1721883}
}
Kostenlos starten →