pp. 2013-2026
S&M4038 Research Paper of Special Issue https://doi.org/10.18494/SAM5469 Published: May 30, 2025 Hybrid Convolutional-gated Recurrent Unit Neural Network Model for Prediction of Weather Indicators [PDF] Hongtao Zhang, Tsungming Lo, Wanying Zhang, Ho Sheng Chen, Jian Xiong, Chihmin Yu, and Tiansyung Lan (Received November 11, 2024; Accepted May 9, 2025) Keywords: weather prediction, convolutional neural network (CNN), gated recurrent unit neural network (GRU), long short-term memory (LSTM), feedforward neural network (FNN)
In this study, we developed a novel integrated model that combines a convolutional neural network (CNN) with a gated recurrent unit neural network (GRU) for the prediction of regional weather. The CNN–GRU model addresses the inherent complexity influenced by global environmental factors in capturing high-level data characteristics and reduces the error rate of current climate index prediction models. The model conducts time-series prediction gathering high-level data characteristics based on CNN and GRU’s capacity. For an accurate prediction, the CNN–GRU model integrates data and retrieved features from it. In the experiment, data including four climatic indices of Beijing, China from 1901 to 2022 were used to construct two-dimensional time-series matrices. The model outperformed the other models including the long short-term memory (LSTM)–fusion neural network, bilateral LSTM, bilateral GRU, double-LSTM, double-GRU, and CNN–GRU (Double-Conv2d) models. The CNN–GRU model was more accurate than the other models.
Corresponding author: Tsungming Lo![]() ![]() This work is licensed under a Creative Commons Attribution 4.0 International License. Cite this article Hongtao Zhang, Tsungming Lo, Wanying Zhang, Ho Sheng Chen, Jian Xiong, Chihmin Yu, and Tiansyung Lan, Hybrid Convolutional-gated Recurrent Unit Neural Network Model for Prediction of Weather Indicators, Sens. Mater., Vol. 37, No. 5, 2025, p. 2013-2026. |