Văn Thị Huỳnh Như , Thân Thị Hương Mai , Ôn Bích Gia Hân Trần Phước Lộc *

* Tác giả liên hệ (tploc@ctu.edu.vn)

Abstract

This study used monthly average rainfall data from the Da Lat meteorological station for the period January 2002 to December 2024 to analyze and forecast rainfall variability. Ten missing observations (3.62%) were imputed using linear interpolation after assessing the missing-data mechanism. Seven forecasting models were evaluated, including four standalone models (Holt–Winters, SARIMA, Prophet, and LSTM) and three residual-based hybrid models (SARIMA+LSTM, HW+LSTM, and Prophet+LSTM). The LSTM component was enhanced through an ensemble strategy with two random seeds, the incorporation of sinusoidal seasonal features (sine/cosine), and a 24-month lookback window. The results showed that the Prophet+LSTM hybrid model achieved the lowest MAE (61.84 mm) among the models considered. In addition, the residuals satisfied the assumptions of white noise and approximate normality, indicating adequate model performance. Forecast results from research models provide a scientific basis for water-resource management and disaster-risk mitigation in the Lam Vien Plateau region.

Keywords: Hybrid model, LSTM, Prophet, rainfall forecasting, SARIMA

Tóm tắt

Trong nghiên cứu này, chuỗi số liệu lượng mưa trung bình tháng tại trạm khí tượng Đà Lạt trong giai đoạn 01/2002–12/2024 đã được sử dụng nhằm đánh giá và dự báo xu thế biến động lượng mưa. Mười giá trị thiếu (3,62%) được xử lý bằng nội suy tuyến tính sau khi kiểm tra cơ chế thiếu dữ liệu. Bảy mô hình dự báo được xem xét, gồm bốn mô hình đơn lẻ (Holt–Winters, SARIMA, Prophet và LSTM) và ba mô hình lai kết hợp với LSTM trên phần dư (SARIMA+LSTM, HW+LSTM và Prophet+LSTM). Mô hình LSTM được cải tiến thông qua kỹ thuật ensemble với hai hạt giống ngẫu nhiên, bổ sung đặc trưng mùa vụ dạng sin/cos và cửa sổ trễ 24 tháng. Kết quả cho thấy Prophet+LSTM đạt sai số MAE thấp nhất (61,84 mm) trong các mô hình khảo sát, đồng thời phần dư thỏa mãn các giả định về nhiễu trắng và phân phối chuẩn. Kết quả dự báo từ các mô hình nghiên cứu cung cấp cơ sở khoa học cho quản lý tài nguyên nước và giảm nhẹ rủi ro thiên tai tại khu vực cao nguyên Lâm Viên.

Từ khóa: Dự báo lượng mưa, LSTM, mô hình lai, Prophet, SARIMA

Article Details

Tài liệu tham khảo

Box, G. E. P., Jenkins, G. M., Reinsel, G. C., & Ljung, G. M. (2015). Time series analysis: Forecasting and control (5th ed.). Wiley.

Can, L. T., & Lộc, T. P. (2024). Dự báo chuỗi thời gian với một số mô hình học máy và ứng dụng. Tạp chí Khoa học Đại học Cần Thơ, 60(Số chuyên đề: Khoa học tự nhiên), 32–39. https://doi.org/10.22144/ctujos.2024.341

Cong, T., Ha, T.-H.-G., Vu, G.-L., Bui, H.-N., Dao, N.-Q.-H., Le, M.-H., Dinh, C.-M., Cong, M.-P., & Doan, Q.-V. (2025). Advancing subseasonal extreme rainfall forecasting in Vietnam using machine learning. SOLA, 21, 273–282. https://doi.org/10.2151/sola.2025-034

Cục Quản lý đê điều và phòng, chống thiên tai. (2024). Công bố số liệu thiệt hại do thiên tai gây ra trong phạm vi cả nước năm 2024. http://phongchongthientai.mard.gov.vn/Pages/cong-bo-so-lieu-thiet-hai-do-thien-tai-gay-ra-trong-pham-vi-ca-nuoc-nam-2024.aspx

Cục Thống kê. (2025). Cơ sở dữ liệu thống kê – Chỉ tiêu lượng mưa trung bình tháng tại trạm khí tượng Đà Lạt (mã dữ liệu: V0109).
https://www.nso.gov.vn/

Diebold, F. X., & Mariano, R. S. (1995). Comparing predictive accuracy. Journal of Business & Economic Statistics, 13(3), 253–263. https://doi.org/10.1080/07350015.1995.10524599

Dimri, T., Ahmad, S., & Sharif, M. (2020). Time series analysis of climate variables using seasonal ARIMA approach. Journal of Earth System Science, 129, Article 149.
https://doi.org/10.1007/s12040-020-01408-x

Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735–1780.
https://doi.org/10.1162/neco.1997.9.8.1735

Holt, C. C. (2004). Forecasting seasonals and trends by exponentially weighted moving averages. International Journal of Forecasting, 20(1), 5–10. https://doi.org/10.1016/j.ijforecast.2003.09.015

IPCC. (2023). Summary for Policymakers. In H. Lee, & J. Romero (Eds.), Climate Change 2023: Synthesis Report. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (pp. 1–34). IPCC.

Kratzert, F., Klotz, D., Brenner, C., Schulz, K., & Herrnegger, M. (2018). Rainfall–runoff modelling using Long Short-Term Memory (LSTM) networks. Hydrology and Earth System Sciences, 22(11), 6005–6022.
https://doi.org/10.5194/hess-22-6005-2018

Huynh, M. V. T., Tran, T. V., Nguyen, N. D. G., Bui, T. B. L., Nguyen, T. T., Ngo, P. C., Meraj, G., Kumar, P., Le, V. T., Doan, V. D., Nguyen, V. T., Downes, N. K., Bhuyan, M. S., Kanga, S., & Singh, S. K. (2024). Modelling and predicting annual rainfall over the Vietnamese Mekong Delta (VMD) using SARIMA. Discover Geoscience, 2(1), 19. https://doi.org/10.1007/s44288-024-00018-0

Moritz, S., & Bartz-Beielstein, T. (2017). imputeTS: Time series missing value imputation in R. The R Journal, 9(1), 207–218.

https://doi.org/10.32614/RJ-2017-009Nguyen, H. D., Pham Van, C., Nguyen, Q.-H., & Bui, Q.-T. (2023). Daily streamflow prediction based on the long short-term memory algorithm: A case study in the Vietnamese Mekong Delta. Journal of Water and Climate Change, 14(4), 1247–1267. https://doi.org/10.2166/wcc.2023.419

Taylor, S. J., & Letham, B. (2018). Forecasting at scale. The American Statistician, 72(1), 37–45. https://doi.org/10.1080/00031305.2017.1380080

Winters, P. R. (1960). Forecasting sales by exponentially weighted moving averages. Management Science, 6(3), 324–342. https://doi.org/10.1287/mnsc.6.3.324

Wu, X., Zhou, J., Yu, H., Liu, D., Xie, K., Chen, Y., Hu, J., Sun, H., & Xing, F. (2021). The development of a hybrid Wavelet-ARIMA-LSTM model for precipitation amounts and drought analysis. Atmosphere, 12(1), 74. https://doi.org/10.3390/atmos12010074

Zhang, G. P. (2003). Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing, 50, 159–175. https://doi.org/10.1016/S0925-2312(01)00702-0