Comparison of XGBoost and LightGBM for 1-Hour-Ahead Significant Wave Height Prediction

Aulia Resty Nur Aini, Rumini Rumini

Abstract


Very short-term significant wave height (Hs) forecasting is important for early warnings related to coastal safety and maritime navigation. This study compares XGBoost and LightGBM for one-hour-ahead Hs prediction using Gold Coast buoy data collected from 2014 to 2024 at 30-minute intervals. Model evaluation employed a rolling-origin validation scheme with three test years (2022, 2023, and 2024). Hyperparameters were independently tuned for each fold using Optuna, with 50 trials per model and equivalent search spaces across both algorithms. The models predicted changes in Hs, which were subsequently converted back to Hs levels. Across the three test years combined, LightGBM achieved an RMSE of 0.0787 m and an R² of 0.131 for Hs changes, whereas XGBoost achieved an RMSE of 0.0790 m and an R² of 0.123. The persistence baseline yielded an RMSE of 0.0844 m. A paired block bootstrap test indicated that the relative RMSE difference of XGBoost compared with LightGBM was −0.49% (95% confidence interval: −0.85% to −0.12%), with negative values indicating better performance by LightGBM. This result remained consistent across block lengths of 48 and 144 time steps. The performance difference was statistically distinguishable in 2022, negligible in 2023, and inconclusive in 2024. Under a narrower hyperparameter search space, the two models performed similarly, suggesting that the observed performance difference was sensitive to hyperparameter tuning. Both models explained only approximately 12–13% of the variance in one-hour Hs changes. Overall, LightGBM was slightly more accurate than XGBoost for one-hour-ahead Hs prediction, although the difference was small, at approximately 0.4 mm.

Keywords


LightGBM; rolling-origin; Significant wave height; very short-term prediction; XGBoost

Full Text:

PDF

References


D. Saydam Irim, M. Sarıkaya, and S. Dağli, “Data-Driven Prediction of Wave Energy Potential in the Black Sea using Ensemble Learning and Real-Time Buoy Observations,” Arab. J. Sci. Eng., Vol. 51, No. 14, pp. 17069–17089, Mar. 2026, DOI: 10.1007/S13369-026-11224-Y.

N. Raj and R. Prakash, “Assessment and Prediction of Significant Wave Height using Hybrid CNN-BiLSTM Deep Learning Model for Sustainable Wave Energy in Australia,” Sustain. Horizons, Vol. 11, p. 100098, Sep. 2024, DOI: 10.1016/J.HORIZ.2024.100098.

I. Yanchin and C. Guedes Soares, “Gradient Boosted Trees and Denoising Autoencoder to Correct Numerical Wave Forecasts,” J. Mar. SCI. Eng., Vol. 12, No. 9, p. 1573, Sep. 2024, DOI: 10.3390/JMSE12091573.

M. O. Costa, R. M. Campos, and C. Guedes Soares, “Enhancing the Accuracy of Metocean Hindcasts with Machine Learning Models,” Ocean Eng., Vol. 287, p. 115724, Nov. 2023, DOI: 10.1016/J.OCEANENG.2023.115724.

C. S. Alfredo and D. Adytia, “Time Series Forecasting of Significant Wave Height using GRU, CNN-GRU, and LSTM,” J. RESTI (Rekayasa Sist. dan Teknol. Informasi), Vol. 6, No. 5, pp. 776–781, Oct. 2022, DOI: 10.29207/RESTI.V6I5.4160.

T. Chen and C. Guestrin, “XGBoost: A Scalable Tree Boosting System,” Proc. ACM SIGKDD Int. Conf. Knowl. Discov. Data Min., pp. 785–794, Aug. 2016, DOI: 10.1145/2939672.2939785.

G. Ke et al., “LightGBM: A Highly Efficient Gradient Boosting Decision Tree,” Adv. Neural Inf. Process. Syst., Vol. 30, pp. 3146–3154, 2017, DOI: 10.5555/3294996.3295074.

F. Anggraeni, D. Adytia, and A. W. Ramadhan, “Forecasting of Wave Height Time Series using AdaBoost and XGBoost, Case Study in Pangandaran, Indonesia,” 2021 Int. Conf. Data SCI. Its Appl. ICoDSA 2021, pp. 97–101, 2021, DOI: 10.1109/ICODSA53588.2021.9617524.

H. Hu, A. J. van der Westhuysen, P. Chu, and A. Fujisaki-Manome, “Predicting Lake Erie Wave Heights and Periods using XGBoost and LSTM,” Ocean Model., Vol. 164, p. 101832, Aug. 2021, DOI: 10.1016/J.OCEMOD.2021.101832.

P. Martina Maria Pushpam and V. S. Felix Enigo, “Forecasting Significant Wave Height using RNN-LSTM Models,” Proc. Int. Conf. Intell. Comput. Control Syst. ICICCS 2020, pp. 1141–1146, May 2020, DOI: 10.1109/ICICCS48265.2020.9121040.

T. Song et al., “Prediction of Significant Wave Height based on EEMD and Deep Learning,” Front. Mar. SCI., Vol. 10, p. 1089357, Feb. 2023, DOI: 10.3389/FMARS.2023.1089357.

E. Androulakis and G. Galanis, “A Two-Step Hybrid System Towards Optimized Wave Height Forecasts,” Stoch. Environ. Res. Risk Assess., Vol. 36, No. 3, pp. 753–766, Aug. 2021, DOI: 10.1007/S00477-021-02075-0.

M. Karbasi, M. Jamei, M. Ali, S. Abdulla, X. Chu, and Z. M. Yaseen, “Developing a Novel Hybrid Auto Encoder Decoder Bidirectional Gated Recurrent Unit model Enhanced with Empirical Wavelet Transform and Boruta-Catboost to Forecast Significant Wave Height,” J. Clean. Prod., Vol. 379, p. 134820, Dec. 2022, DOI: 10.1016/J.JCLEPRO.2022.134820.

J. Shi et al., “A Machine-Learning Approach based on Attention Mechanism for Significant Wave Height Forecasting,” J. Mar. SCI. Eng., Vol. 11, No. 9, p. 1821, Sep. 2023, DOI: 10.3390/JMSE11091821.

Y. Zhan, H. Zhang, J. Li, and G. Li, “Prediction Method for Ocean Wave Height based on Stacking Ensemble Learning Model,” J. Mar. SCI. Eng., Vol. 10, No. 8, p. 1150, Aug. 2022, DOI: 10.3390/JMSE10081150.

S. Gracia, J. Olivito, J. Resano, B. Martin-del-Brio, M. de Alfonso, and E. Álvarez, “Improving Accuracy on Wave Height Estimation through Machine Learning Techniques,” Ocean Eng., Vol. 236, p. 108699, Sep. 2021, DOI: 10.1016/J.OCEANENG.2021.108699.

M. S. Afzal, L. Kumar, V. Chugh, Y. Kumar, and M. Zuhair, “Prediction of Significant Wave Height using Machine Learning and Its Application to Extreme Wave Analysis,” J. Earth Syst. SCI., Vol. 132, No. 2, p. 51, Mar. 2023, DOI: 10.1007/S12040-023-02058-5.

M. Ré Henriques, D. Silva, I. Yanchin, M. Latas, and C. Guedes Soares, “Improving the Forecast of Wind Speed and Significant Wave Height using Neural Networks and Gradient Boosting Trees,” Ocean Eng., Vol. 327, p. 120925, May 2025, DOI: 10.1016/J.OCEANENG.2025.120925.

D. Adytia, D. Saepudin, S. R. Pudjaprasetya, S. Husrin, and A. Sopaheluwakan, “A Deep Learning Approach for Wave Forecasting based on a Spatially Correlated Wind Feature, with a Case Study in the Java Sea, Indonesia,” Fluids, Vol. 7, No. 1, p. 39, Jan. 2022, DOI: 10.3390/FLUIDS7010039.

A. Hasan, I. Kayes, M. Alam, T. Shahriar, and M. Ahsan Habib, “Generalized Machine Learning Models to Predict Significant Wave Height Utilizing Wind and Atmospheric Parameters,” Energy Convers. Manag. X, Vol. 23, p. 100623, Jul. 2024, DOI: 10.1016/J.ECMX.2024.100623.

A. A. M. Ahmed et al., “Hybrid Deep Learning Model for Wave Height Prediction in Australia’s Wave Energy Region,” Appl. Soft Comput., Vol. 150, Jan. 2024, DOI: 10.1016/J.ASOC.2023.111003.

Rohitgds, “Gold Coast City Wave Dataset from 2014-2024,” Kaggle, 2025. Accessed: Jun. 23, 2026. [Online]. Available: https://www.kaggle.com/datasets/rohitgds/gold-coast-city-wave-dataset-from-2014-2024

H. Hewamalage, K. Ackermann, and C. Bergmeir, “Forecast Evaluation for Data Scientists: Common Pitfalls and Best Practices,” Data Min. Knowl. Discov., Vol. 37, No. 2, pp. 788–832, 2023, DOI: 10.1007/S10618-022-00894-5.

T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Koyama, “Optuna: A Next-Generation Hyperparameter Optimization Framework,” Proc. ACM SIGKDD Int. Conf. Knowl. Discov. Data Min., pp. 2623–2631, Jul. 2019, DOI: 10.1145/3292500.3330701.

R. Lunde and C. Shalizi, “Bootstrapping Generalization Error Bounds for Time Series,” Sankhya A, pp. 1–34, Aug. 2026, DOI: 10.1007/S13171-026-00452-X.

S. M. Lundberg and S.-I. Lee, “A Unified Approach to Interpreting Model Predictions,” Adv. Neural Inf. Process. Syst., Vol. 30, pp. 4765–4774, 2017, DOI: 10.48550/arXiv.1705.07874.




DOI: https://doi.org/10.32520/stmsi.v15i11.7066

Article Metrics

Abstract view : 0 times
PDF - 0 times

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.