Enhancing Load Forecasting Accuracy of Neural Hierarchical Interpolation for Time Series through Hyperparameter Optimization
Proceedings of Engineering and Technology Innovation · 26.06.2026 · 34 · 17-32
DOI: 10.46604/peti.2026.16317 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of Engineering and Technology Innovation, online erschienen am 26.06.2026. Verfasst von Thanh Ngoc Tran und 1 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
This study aims to improve the forecasting accuracy of the neural hierarchical interpolation for time series (NHITS) model through a hyperparameter optimization framework. Three optimization strategies, namely random search (RS), tree-structured Parzen estimator (TPE), and genetic algorithm (GA), are compared under the same search space and evaluation budget. Three key hyperparameters, including mlp_units, learning_rate, and max_steps, are optimized because they directly affect model capacity, convergence behavior, and training effort. Experiments are conducted on two Australian half-hourly electricity demand datasets, New South Wales (NSW) and Queensland (QLD), each containing 122,735 records. Forecasting performance is evaluated using MSE, RMSE, MAE, and MAPE. The results show that all optimization methods improve the default NHITS configuration, while TPE achieves the best performance. MAPE decreases from 2.21% to 1.23% for NSW and from 3.34% to 1.79% for QLD.
Zitieren
APA
Tran, T. N., & Nguyen, T. A. (2026). Enhancing Load Forecasting Accuracy of Neural Hierarchical Interpolation for Time Series through Hyperparameter Optimization. Proceedings of Engineering and Technology Innovation, 34, 17-32. https://doi.org/10.46604/peti.2026.16317
Vancouver
Tran TN, Nguyen TA. Enhancing Load Forecasting Accuracy of Neural Hierarchical Interpolation for Time Series through Hyperparameter Optimization. Proceedings of Engineering and Technology Innovation. 2026;34:17-32. doi:10.46604/peti.2026.16317
BibTeX
@article{Tran2026,
author = {Tran, Thanh Ngoc and Nguyen, Tuan Anh},
title = {Enhancing Load Forecasting Accuracy of Neural Hierarchical Interpolation for Time Series through Hyperparameter Optimization},
journal = {Proceedings of Engineering and Technology Innovation},
year = {2026},
volume = {34},
pages = {17--32},
doi = {10.46604/peti.2026.16317}
}
EUVIMED