@article{4599,
  author = {Wei Xiao},
  title = {Liquid Quality Prediction Using Gated Recurrent Unit (GRU) Networks: A Comparative Study with LSTM and RNN Models},
  journal = {Journal of Information & Systems Management},
  year = {2025},
  volume = {15},
  number = {4},
  doi = {https://doi.org/10.6025/jism/2025/15/4/153-160},
  url = {https://www.dline.info/jism/fulltext/v15n4/jismv15n4_1.pdf},
  abstract = {The paper investigates water quality prediction using machine learning, with a focus on the Gated Recurrent
Unit (GRU) model. As water pollution intensifies due to industrialization and urbanization, accurate forecasting
becomes critical for environmental protection and sustainable water resource management. The study reviews
existing approaches, including mechanistic models like Streeter Phelps and WASP, and data driven non
mechanistic models such as ARIMA, SVR, and deep learning methods. Among deep learning architectures,
GRU a simplified variant of LSTM offers strong performance in capturing long term temporal dependencies in
time series data while maintaining lower computational complexity. The authors construct a GRU based
water quality prediction model using monthly pH and dissolved oxygen (DO) data from China's Surface Water
Quality Automatic Monitoring system (2011-2018). After preprocessing and normalization, the GRU
model is trained on data from 2011-2017 and tested on 2018 data. Results show that GRU outperforms
traditional RNNs and LSTMs in prediction accuracy for both pH and DO levels. The study concludes that GRU
provides an efficient and precise solution for water quality forecasting, thereby supporting better decisionmaking
in environmental management. Future work may focus on optimizing model architecture and feature
selection to enhance robustness and generalization further. This research highlights the potential of deep
learning, particularly GRU networks, in tackling real world environmental challenges.},
}