AutoML 기반 단기 해상상태 예측 모델 연구

서강현¹ ; 송유정²^{, *}

1해군사관학교 인공지능학과 조교수
2순천향대학교 컴퓨터소프트웨어공학과 조교수

Research on Short-term Ocean State Prediction Using AutoML

Kanghyeon Seo¹ ; Yoojeong Song²^{, *}

1ROK Navy/Assistant Professor, Dept. of Artificial Intelligence, Republic of Korea Naval Academy
2Assistant professor, Dept. of Computer Software Engineering, Soonchunhyang University

Correspondence to: *Yoojeong Song Dept. of Computer Software Engineering, Soonchunhyang University 22, Soonchunhyang-ro, Asan-si, Chungcheongnam-do 31538, Republic of Korea Tel: +82-41-530-1350 E-mail: yoojeong@sch.ac.kr

초록

본 연구에서는 AutoML(자동화된 머신러닝) 기술을 해양 상태 예측에 적용한다. 특히, AutoGluon 라이브러리를 활용하여 전문적인 머신러닝 지식 없이도 복잡한 시계열 예측 모델을 신속하게 개발하는 방법론을 제시한다. 제공된 해양 관측 부이의 시계열 데이터를 사용하여 주요 해상 상태 지표인 최대파고를 예측하는 모델을 구축한다. 실증적인 연구 결과 AutoML 기반 모델은 최소한의 개발 노력으로도 우수한 예측 성능을 보인다. 본 연구는 해군 작전 환경에서 요구되는 정확하고 신속한 의사결정을 지원하기 위해, 데이터 기반의 해상 상태 예측 시스템을 효율적으로 구축하고 운용할 수 있는 실용적인 방안을 제시한다는 점에서 의의가 있다.

Abstract

This study applies Automated Machine Learning (AutoML) technology to ocean condition prediction. Specifically, it presents a methodology for rapidly developing complex time-series forecasting models without specialized machine learning expertise by utilizing the AutoGluon library. A model for predicting maximum wave height, a key marine state indicator, is constructed using time-series data from ocean observation buoys. Empirical research results demonstrate that the AutoML-based model achieves superior predictive performance with minimal development effort. This study is significant in that it proposes a practical approach to efficiently build and operate a data-driven ocean condition prediction system to support the accurate and rapid decision-making required in naval operational environments.

Keywords:

Automated Machine Learning, Ocean State Prediction, Time-series Forecasting, Deep Learning, Machine Learning, AutoGluon

키워드:

자동화 머신러닝, 해양상태예측, 시계열 예측, 딥러닝, 머신러닝, 오토글루온

References

Booij, N., Ris, R. C., & Holthuijsen, L. H. (1999). “A Third-generation Wave Model for Coastal Regions: 1. Model Description and Validation.” Journal of Geophysical Research: Oceans, Vol. 104, No. C4, pp. 7649-7666.
Tolman, H. L. (1991). “A Third-generation Model for Wind Waves on Slowly Varying, Unsteady, and Inhomogeneous Depths and Currents.” Journal of Physical Oceanography, Vol. 21, No. 6, 782-797.
He, Xin, Kaiyong Zhao, and Xiaowen Chu. (2021). “AutoML: A Survey of the StIte-of-the-art.” Knowledge-based Systems, Vol. 212, 106622.
Komen, Gerbrand Johan, et al. (1994). Dynamics and Modelling of Ocean Waves. Vol. 532. UK: Cambridge University Press.
Agrawal, J. D., and M. C. Deo. (2002). “On-line Wave Prediction.” Marine Structures, Vol. 15, No. 1, pp. 57-74.
Deo, M. C., and C. Sridhar Naidu. (1998). “Real Time Wave Forecasting Using Neural Networks.” Ocean Engineering, Vol. 26, No. 3, pp. 191-203.
Gao, Meng, et al. (2021). “Efficient Multi-angle Polarimetric Inversion of Aerosols and Ocean Color Powered by a Deep Neural Network Forward Model.” Atmospheric Measurement Techniques, Vol. 14, No. 6, pp. 4083-4110.
Vaswani, Ashish, et al. (2017). “Attention is All You Need.” Advances in Neural Information Processing Systems 30.
Erickson, Nick, et al. “Autogluon-tabular: Robust and Accurate AutoML for Structured Data.” arXiv preprint arXiv:2003.06505, (2020).
기상청, 기상개방포털, https://data.kma.go.kr/cmmn/main.do