Journal of Korean Society of Coastal and Ocean Engineers (한국해안·해양공학회논문집)

Korean Society of Coastal and Ocean Engineers (한국해안·해양공학회)
권호 표지
DOI QR코드

DOI QR Code

Imputation of Missing SST Observation Data Using Multivariate Bidirectional RNN

다변수 Bidirectional RNN을 이용한 표층수온 결측 데이터 보간

  • Shin, YongTak (Department of Ocean Sciences, College of Natural Science, Inha University) ;
  • Kim, Dong-Hoon (Artificial Intelligence Convergence Research Center, Inha University) ;
  • Kim, Hyeon-Jae (Department of Ocean Sciences, College of Natural Science, Inha University) ;
  • Lim, Chaewook (Department of Ocean Sciences, College of Natural Science, Inha University) ;
  • Woo, Seung-Buhm (Department of Ocean Sciences, College of Natural Science, Inha University)
  • 신용탁 (인하대학교 해양과학과) ;
  • 김동훈 (인하대학교 인공지능융합센터) ;
  • 김현재 (인하대학교 해양과학과) ;
  • 임채욱 (인하대학교 해양과학과) ;
  • 우승범 (인하대학교 해양과학과)
  • Received : 2022.07.05
  • Accepted : 2022.08.19
  • Published : 2022.08.31
Download PDF
⟨ Previous Next ⟩

Abstract

The data of the missing section among the vertex surface sea temperature observation data was imputed using the Bidirectional Recurrent Neural Network(BiRNN). Among artificial intelligence techniques, Recurrent Neural Networks (RNNs), which are commonly used for time series data, only estimate in the direction of time flow or in the reverse direction to the missing estimation position, so the estimation performance is poor in the long-term missing section. On the other hand, in this study, estimation performance can be improved even for long-term missing data by estimating in both directions before and after the missing section. Also, by using all available data around the observation point (sea surface temperature, temperature, wind field, atmospheric pressure, humidity), the imputation performance was further improved by estimating the imputation data from these correlations together. For performance verification, a statistical model, Multivariate Imputation by Chained Equations (MICE), a machine learning-based Random Forest model, and an RNN model using Long Short-Term Memory (LSTM) were compared. For imputation of long-term missing for 7 days, the average accuracy of the BiRNN/statistical models is 70.8%/61.2%, respectively, and the average error is 0.28 degrees/0.44 degrees, respectively, so the BiRNN model performs better than other models. By applying a temporal decay factor representing the missing pattern, it is judged that the BiRNN technique has better imputation performance than the existing method as the missing section becomes longer.

정점 표층 수온 관측 데이터 중 결측 구간의 데이터를 양방향 순환신경망(Bidirectional Recurrent Neural Network, BiRNN) 기법을 이용하여 보간하였다. 인공지능 기법 중 시계열 데이터에 일반적으로 활용되는 Recurrent Neural Networks(RNNs)은 결측 추정 위치까지의 시간 흐름 방향 또는 역방향으로만 추정하기 때문에 장기 결측 구간에는 추정 성능이 떨어진다. 반면, 본 연구에서는 결측 구간 전후의 양방향으로 추정을 하여 장기 결측 데이터에 대해서도 추정 성능을 높일 수 있다. 또한 관측점 주위의 가용한 모든 데이터(수온, 기온, 바람장, 기압, 습도)를 사용함으로써, 이들 상관관계로부터 보간 데이터를 함께 추정하도록 하여 보간 성능을 더욱 높이고자 하였다. 성능 검증을 위하여 통계 기반 모델인 Multivariate Imputation by Chained Equations(MICE)와 기계학습 기반의 Random Forest 모델, 그리고 Long Short-Term Memory(LSTM)을 이용한 RNN 모델과 비교하였다. 7일간의 장기 결측에 대한 보간에 대해서 BiRNN/통계 모델들의 평균 정확도가 각각 70.8%/61.2%이며 평균 오차가 각각 0.28도/0.44도로 BiRNN 모델이 다른 모델보다 좋은 성능을 보인다. 결측 패턴을 나타내는 temporal decay factor를 적용함으로써 BiRNN 기법이 결측 구간이 길어질수록 보간 성능이 기존 방법보다 우수한 것으로 판단된다.

Keywords

Acknowledgement

이 논문은 2022년도 해양수산부 재원으로 해양수산과학기술진흥원의 지원을 받아 수행된 연구임(20220051, 경기·인천 씨그랜트). 이 논문은 2022년도 정부(과학기술정보통신부)의 재원으로 정보통신기획평가원의 지원을 받아 수행된 연구임(2020-0-01389, 인공지능융합연구센터지원(인하대학교)).

References

  1. Azur, M.J., Stuart, E.A., Frangakis, C. and Leaf, P.J. (2011). Multiple imputation by chained equations: what is it and how does it work?. International Journal of Methods in Psychiatric Research, 20(1), 40-49. https://doi.org/10.1002/mpr.329
  2. Bengio, S., Vinyals, O., Jaitly, N. and Shazeer, N. (2015). Scheduled sampling for sequence prediction with recurrent neural networks. Advances in Neural Information Processing Systems, 28.
  3. Cao, W., Wang, D., Li, J., Zhou, H., Li, L. and Li, Y. (2018). Brits: Bidirectional recurrent imputation for time series. Advances in Neural Information Processing Systems, 31.
  4. Cho, H.Y., Jeong, J.Y., Shim, J.S. and Kim, S.J. (2010). Variation pattern analysis on the air and surface water temperatures of the yellow sea monitoring buoy. Journal of Korean Society of Coastal and Ocean Engineers, 22(5), 316-325 (in Korean).
  5. Feng, T. and Narayanan, S. (2019). Imputing missing data in large-scale multivariate biomedical wearable recordings using bidirectional recurrent neural networks with temporal activation regularization. In 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) (pp. 2529-2534). IEEE.
  6. Junger, W.L. and De Leon, A.P. (2015). Imputation of missing data in time series for air pollutants. Atmospheric Environment, 102, 96-104. https://doi.org/10.1016/j.atmosenv.2014.11.049
  7. Kim, H.J., Kim, D.H., Lim, C.W., Shin, Y.T., Lee, S.C., Choi, Y.J. and Woo, S.B. (2021). An outlier detection using autoencoder for ocean observation data. Journal of Korean Society of Coastal and Ocean Engineers, 33(6), 265-274 (in Korean). https://doi.org/10.9765/KSCOE.2021.33.6.265
  8. Kim, Y.J. and Chi, M. (2018). Temporal Belief Memory: Imputing Missing Data during RNN Training. In Proceedings of the 27th International Joint Conference on Artificial Intelligence (IJCAI-2018).
  9. Lipton, Z.C., Kale, D.C. and Wetzel, R. (2016). Modeling missing data in clinical time series with rnns. Machine Learning for Healthcare, 56, 253-270.
  10. McNeil, N. and Chirtkiatsakul, B. (2016). Statistical models for the pattern of sea surface temperature in the North Atlantic during 1973-2008. International Journal of Climatology, 36(11), 3856-3863. https://doi.org/10.1002/joc.4598
  11. Miller, D. and Kim, J.M. (2021). Univariate and multivariate machine learning forecasting models on the price returns of cryptocurrencies. Journal of Risk and Financial Management, 14(10), 486.
  12. Mohebzadeh, H., Mokari, E., Daggupati, P. and Biswas, A. (2021). A machine learning approach for spatiotemporal imputation of MODIS chlorophyll-a. International Journal of Remote Sensing, 42(19), 7381-7404. https://doi.org/10.1080/01431161.2021.1957513
  13. Qu, B., Gabric, A.J., Zhu, J.N., Lin, D.R., Qian, F. and Zhao, M. (2012). Correlation between sea surface temperature and wind speed in Greenland Sea and their relationships with NAO variability. Water Science and Engineering, 5(3), 304-315.
  14. Suo, Q., Yao, L., Xun, G., Sun, J. and Zhang, A. (2019). Recurrent imputation for multivariate time series with missing values. In 2019 IEEE International Conference on Healthcare Informatics (ICHI) (pp. 1-3). IEEE.
  15. Tang, F. and Ishwaran, H. (2017). Random forest missing data algorithms. Statistical Analysis and Data Mining: The ASA Data Science Journal, 10(6), 363-377. https://doi.org/10.1002/sam.11348
  16. Van Buuren, S. and Oudshoorn, K. (1999). Flexible multivariate imputation by MICE (pp. 1-20). Leiden: TNO.
  17. Worku, G., Teferi, E., Bantider, A., Dile, Y.T. and Taye, M.T. (2018). Evaluation of regional climate models performance in simulating rainfall climatology of Jemma sub-basin, Upper Blue Nile Basin, Ethiopia. Dynamics of Atmospheres and Oceans, 83, 53-63. https://doi.org/10.1016/j.dynatmoce.2018.06.002
  18. Yang, C.H., Wu, C.H., Hsieh, C.M., Wang, Y.C., Tsen, I.F. and Tseng, S.H. (2021). Deep learning for imputation and forecasting tidal level. IEEE Journal of Oceanic Engineering, 46(4), 1261-1271. https://doi.org/10.1109/JOE.2021.3073931