海岸遊憩活動已成為民眾的日常休閒，然而，海岸瘋狗浪(Coastal Freak Wave)時常無預警的出現，將岸邊人員捲入海中，其形成原因尚未擁有完整的理論進行描述，使得預測海岸瘋狗浪的發生時機仍是一個挑戰。現今電腦技術演進與計算能力的增強促進了人工智慧(Artificial Intelligence, AI)的蓬勃發展，能夠處理多變量、非線性及大量的資料，為理解複雜的自然現象提供了重要協助。因此本研究使用人工智慧中具有時序列處理能力的長短期記憶網路(Long Short-term Memory, LSTM)建立海岸瘋狗浪的預測模型，除了探討長短期記憶網路的模型建置結果與預測特性外，也會與其他不具備時序列分析功能的人工智慧方法進行比較。
本研究透過光學攝影機取得海岸瘋狗浪發生時間，並蒐集鄰近浮標的海氣象數據建立模型的資料庫。利用長短期記憶網路分析過去一段時間內的歷史海氣象數據，以預測海岸瘋狗浪的發生。研究結果顯示模型的預測表現良好，整體評估指標可達八成以上，無論實際是否有發生海岸瘋狗浪，模型都具有穩定的預測能力。
本文建立之模型與其他人工智慧模型相比的結果顯示，無論何種模型均具有正確預測海岸瘋狗浪的能力，各評估指標介於七至八成之間。然而在長浪期間，本文建立之模型預測的發生機率明顯高於其他人工智慧模型。在浪況較不穩定的颱風海上警報期間的預測結果也顯示長短期記憶網路模型的平均預測準確率達80%，比其他人工智慧預測模型高出約10%，平均預測精確率可達87.3%，明顯優於其他預測模型，表示長短期記憶網路模型的誤報率較低。
本研究結果顯示長短期記憶網路能夠利用過去一段時間內的海氣象數據成功預測海岸瘋狗浪，顯示應用此類人工智慧技術在預測海岸瘋狗浪方面具有可行性與可靠性。

Coastal freak waves, which often appear unexpectedly at coastlines without warning, pose a significant threat to individuals near the shore by pulling them into the sea. The mechanism behind the formation of coastal freak waves lacks a complete theoretical description, making their prediction a major challenge. In recent years, Artificial Intelligence (AI) has emerged as a powerful tool capable of processing large volumes of multivariate and nonlinear data, providing robust solutions for understanding complex natural phenomena. This study utilizes Long Short-Term Memory (LSTM), a type of AI with sequential data processing capabilities, to establish a predictive model for coastal freak waves. In addition to discussing the model construction results and predictive characteristics of LSTM, this study compares LSTM with other AI methods that do not possess sequential analysis capabilities. The research uses optical cameras to capture the occurrence times of coastal freak waves and collects marine meteorological data from nearby buoys to establish the model's database. Historical marine meteorological data over a period had been analyzed in this study by LSTM to predict the occurrence of these waves. The results demonstrate that the LSTM model performs well, with evaluation metrics reaching 80%, consistently predicting the occurrence of coastal freak waves regardless of actual events. The study also indicates that LSTM outperforms other AI models in predicting coastal freak waves. During periods of high waves, the LSTM model shows significantly higher predictive probabilities compared to other models. During typhoon warnings at sea, the LSTM model achieves an accuracy of 80%, which is approximately 10% higher than other models, with a precision of 87.3%, indicating a lower false alarm rate. Overall, this research confirms that LSTM effectively utilizes historical marine meteorological data to successfully predict coastal freak waves, demonstrating the feasibility and reliability of this technology in predicting such occurrences.

[1]王敘民、陳盈智、董東璟、滕春慈(2017)，海岸瘋狗浪光學影像分析之研究，第39屆海洋工程研討會，第555-560頁。
[2]許明光、曾俊超、高家俊(1993)，臺灣地區「瘋狗浪」之調查及成因初探，第十五屆海洋工程研討會論文集，第513-524頁。
[3]陳威成(2021)，人工智慧(AI)演算法在瘋狗浪機率預警系統建置之研究，國立成功大學碩士論文。
[4]陳威成、陳盈智、林芳如、董東璟、蔡政翰(2022)，隨機森林機器學習法應用於瘋狗浪機率預警系統之研究，第44屆海洋工程研討會，第505-510頁。
[5]陳威成、陳盈智、董東璟 (2020)，支撐向量機應用於瘋狗浪預警系統，第42屆海洋工程研討會，第297-301頁。
[6]陳威成、楊文榮、傅仲偉、董東璟、廖建明(2022)，人工智慧技術在海洋觀測之應用，中國土木水利工程學刊，第四十九卷，第六期，第31-39頁。
[7]陳盈智、董東璟、林芳如、滕春慈(2022)，海岸異常波浪觀測之研究，中國土木水利工程學刊，第四十九卷，第六期，第40-45頁。
[8]Akhmediev, N., & Pelinovsky, E. (2010). Editorial – Introductory remarks on “Discussion & Debate: Rogue Waves – Towards a Unifying Concept?” The European Physical Journal Special Topics, 185(1), 1–4.
[9]Amari, S. (1993). Backpropagation and stochastic gradient descent method. Neurocomputing, 5(4), 185–196.
[10]Azevedo, L., Meyers, S., Pleskachevsky, A., Pereira, H. P. P., & Luther, M. (2022). Characterizing Rogue Waves at the Entrance of Tampa Bay (Florida, USA). Journal of Marine Science and Engineering, 10(4), Article 4.
[11]Chang, Z., Zhang, Y., & Chen, W. (2018). Effective Adam-Optimized LSTM Neural Network for Electricity Price Forecasting. 2018 IEEE 9th International Conference on Software Engineering and Service Science (ICSESS), 245–248.
[12]Chen, C.-S., Cheng, M.-Y., & Wu, Y.-W. (2012). Seismic assessment of school buildings in Taiwan using the evolutionary support vector machine inference system. Expert Systems with Applications, 39(4), 4102–4110.
[13]Chen, W., Pourghasemi, H. R., Kornejady, A., & Zhang, N. (2017). Landslide spatial modeling: Introducing new ensembles of ANN, MaxEnt, and SVM machine learning techniques. Geoderma, 305, 314–327.
[14]Chien, H., Kao, C.-C., & Chuang, L. Z. H. (2002). On the Characteristics of Observed Coastal Freak Waves. Coastal Engineering Journal, 44(4), 301–319.
[15]Choi, C., Kim, J., Kim, J., Kim, D., Bae, Y., & Kim, H. S. (2018). Development of Heavy Rain Damage Prediction Model Using Machine Learning Based on Big Data. Advances in meteorology, 2018.
[16]Doong, D.-J., Chen, S.-T., Chen, Y.-C., & Tsai, C.-H. (2020). Operational Probabilistic Forecasting of Coastal Freak Waves by Using an Artificial Neural Network. Journal of Marine Science and Engineering, 8(3), Article 3.
[17]Dysthe K., Krogstad H. E., & Müller P. (2008). Oceanic Rogue Waves. Annual Review of Fluid Mechanics, 40(Volume 40, 2008), 287–310.
[18]Elman, J. L. (1990). Finding Structure in Time. Cognitive Science, 14(2), 179–211.
[19]Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115–118.
[20]Gao, S., Huang, Y., Zhang, S., Han, J., Wang, G., Zhang, M., & Lin, Q. (2020). Short-term runoff prediction with GRU and LSTM networks without requiring time step optimization during sample generation. Journal of Hydrology, 589, 125188.
[21]Hinton, G., Srivastava, N., & Swersky, K. (2012). Neural networks for machine learning lecture 6a overview of mini-batch gradient descent. Cited on, 14(8), 2.
[22]Hochreiter, S., & Schmidhuber, J. (1997). Long Short-term Memory. Neural Computation, 9(8), 1735–1780.
[23]Hou, J., Wang, Y., Zhou, J., & Tian, Q. (2022). Prediction of hourly air temperature based on CNN–LSTM. Geomatics, Natural Hazards and Risk, 13(1), 1962–1986.
[24]Janssen, P. A. E. M. (2003). Nonlinear Four-Wave Interactions and Freak Waves. Journal of Physical Oceanography, 33(4), 863–884.
[25]Jaseena, K. U., & Kovoor, B. C. (2021). Decomposition-based hybrid wind speed forecasting model using deep bidirectional LSTM networks. Energy Conversion and Management, 234, 113944.
[26]Jörges, C., Berkenbrink, C., & Stumpe, B. (2021). Prediction and reconstruction of ocean wave heights based on bathymetric data using LSTM neural networks. Ocean Engineering, 232, 109046.
[27]Kharif, C., & Pelinovsky, E. (2003). Physical mechanisms of the rogue wave phenomenon. European Journal of Mechanics - B/Fluids, 22(6), 603–634.
[28]Kingma, D. P., & Ba, J. (2014). Adam: A Method for Stochastic Optimization. CoRR.
[29]Lee, J., Kim, C.-G., Lee, J. E., Kim, N. W., & Kim, H. (2018). Application of Artificial Neural Networks to Rainfall Forecasting in the Geum River Basin, Korea. Water, 10(10), Article 10.
[30]Liu, J., Zhang, J., Billah, M. M., & Zhang, T. (2023). ABiLSTM Based Prediction Model for AUV Trajectory. Journal of Marine Science and Engineering, 11(7), Article 7.
[31]Luo, Q.-R., Xu, H., & Bai, L.-H. (2022). Prediction of significant wave height in hurricane area of the Atlantic Ocean using the Bi-LSTM with attention model. Ocean Engineering, 266, 112747.
[32]Máté, D., Raza, H., & Ahmad, I. (2023). Comparative Analysis of Machine Learning Models for Bankruptcy Prediction in the Context of Pakistani Companies. Risks, 11(10), Article 10.
[33]McCulloch, W. S., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. The Bulletin of Mathematical Biophysics, 5(4), 115–133.
[34]Mehtab, S., Sen, J., & Dutta, A. (2021). Stock Price Prediction Using Machine Learning and LSTM-Based Deep Learning Models. In S. M. Thampi, S. Piramuthu, K.-C. Li, S. Berretti, M. Wozniak, & D. Singh (Eds.), Machine Learning and Metaheuristics Algorithms, and Applications (pp. 88–106). Springer.
[35]Meng, Z.-F., Chen, Z., Khoo, B. C., & Zhang, A.-M. (2022). Long-time prediction of sea wave trains by LSTM machine learning method. Ocean Engineering, 262, 112213.
[36]Miotto, R., Li, L., Kidd, B. A., & Dudley, J. T. (2016). Deep Patient: An Unsupervised Representation to Predict the Future of Patients from the Electronic Health Records. Scientific Reports, 6(1), 26094.
[37]Mori, N., & Janssen, P. A. E. M. (2006). On Kurtosis and Occurrence Probability of Freak Waves. Journal of Physical Oceanography, 36(7), 1471–1483.
[38]Okewu, E., Adewole, P., & Sennaike, O. (2019). Experimental Comparison of Stochastic Optimizers in Deep Learning. In S. Misra, O. Gervasi, B. Murgante, E. Stankova, V. Korkhov, C. Torre, A. M. A. C. Rocha, D. Taniar, B. O. Apduhan, & E. Tarantino (Eds.), Computational Science and Its Applications – ICCSA 2019 (pp. 704–715). Springer International Publishing.
[39]Onorato, M., Residori, S., Bortolozzo, U., Montina, A., & Arecchi, F. T. (2013). Rogue waves and their generating mechanisms in different physical contexts. Physics Reports, 528(2), 47–89.
[40]Orzech, M. D., & Wang, D. (2020). Measured Rogue Waves and Their Environment. Journal of Marine Science and Engineering, 8(11), Article 11.
[41]Ruban, V. P. (2007). Nonlinear Stage of the Benjamin-Feir Instability: Three-Dimensional Coherent Structures and Rogue Waves. Physical Review Letters, 99(4), 044502.
[42]Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323(6088), Article 6088.
[43]Sareen, K., Panigrahi, B. K., Shikhola, T., & Nagdeve, R. (2023). An integrated decomposition algorithm based bidirectional LSTM neural network approach for predicting ocean wave height and ocean wave energy. Ocean Engineering, 281, 114852.
[44]ŞEN, S. Y., & ÖZKURT, N. (2020). Convolutional Neural Network Hyperparameter Tuning with Adam Optimizer for ECG Classification. 2020 Innovations in Intelligent Systems and Applications Conference (ASYU), 1–6.
[45]Soomere, T. (2010). Rogue waves in shallow water. The European Physical Journal Special Topics, 185(1), 81–96.
[46]Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. (2014). Dropout: A simple way to prevent neural networks from overfitting. The Journal of Machine Learning Research, 15(1), 1929–1958.
[47]Tamura, H., Waseda, T., & Miyazawa, Y. (2009). Freakish sea state and swell-windsea coupling: Numerical study of the Suwa-Maru incident. Geophysical Research Letters, 36(1).
[48]Tibshirani, R. (1996). Regression Shrinkage and Selection via the Lasso. Journal of the Royal Statistical Society. Series B (Methodological), 58(1), 267–288.
[49]Tsai, C.-H., Su, M.-Y., & Huang, S.-J. (2004). Observations and conditions for occurrence of dangerous coastal waves. Ocean Engineering, 31(5), 745–760.
[50]Wang, C., Venkatesh, S., & Judd, J. (1993). Optimal Stopping and Effective Machine Complexity in Learning. Advances in Neural Information Processing Systems, 6.
[51]Wang, Q., Guo, Y., Yu, L., & Li, P. (2020). Earthquake Prediction Based on Spatio-Temporal Data Mining: An LSTM Network Approach. IEEE Transactions on Emerging Topics in Computing, 8(1), 148–158.
[52]Wei, L., Guan, L., Qu, L., & Guo, D. (2020). Prediction of Sea Surface Temperature in the China Seas Based on Long Short-term Memory Neural Networks. Remote Sensing, 12(17), Article 17.
[53]White, B. S., & Fornberg, B. (1998). On the chance of freak waves at sea. Journal of Fluid Mechanics, 355, 113–138.
[54]Yang, C.-C., & Hwang, C.-H. (2010). Studies on typhoon-wave prediction by Artificial Neural Networks at Longdong, Taiwan. 2010 Sixth International Conference on Natural Computation, 2, 626–630.