Sign in Subscribe
Technical institutes

Deep Learning Model Predicts Atmospheric Ozone Concentrations

Strategic Insights

— 2 min read

A team of researchers from the Xi'an University of Technology has developed a deep learning model that can accurately predict atmospheric ozone concentrations. The model combines an attention mechanism with a convolutional neural network (CNN) and long short-term memory (LSTM) network to address the nonlinear nature of multivariate time-series data. The researchers employed the model to analyze 16,806 records of eight meteorological and environmental parameters from 2018-2019, and the results demonstrated that the model outperformed individual models and the CNN-LSTM model, especially in forward prediction with a multi-hour time lag.

Key Takeaways:

  • The researchers developed a deep learning model that combines an attention mechanism with a CNN and LSTM network to predict atmospheric ozone concentrations.
  • The model takes eight meteorological and environmental parameters from 16,806 records (2018-2019) as input, selected using principal component analysis (PCA).
  • The model features an attention-based CNN-LSTM hybrid deep learning model with specific settings, including a time step of 5, a batch size of 25, 15 units in the LSTM layer, the ReLU activation function, 25 epochs, and an overfitting avoidance strategy with a dropout rate of 0.15.
  • The experimental results demonstrate that the hybrid model outperforms individual models and the CNN-LSTM model, especially in forward prediction with a multi-hour time lag.
  • The model exhibits a high coefficient of determination (R = 0.971) and a root mean square error of 3.59 for a 1-hour time lag.
  • The model also exhibits consistent accuracy across different seasons, highlighting its robustness and superior time-series prediction capabilities for ozone concentrations.
  • The researchers employed the model to analyze data from the Centre for Lidar Remote Sensing Research at the Xi'an University of Technology.
  • The additional authors for the research include Jiang Yuan, Wang Yufeng, Yang Xueting, Di Huige, and Yan Qing.

Statistics:

  • The researchers analyzed 16,806 records of eight meteorological and environmental parameters from 2018-2019.
  • The model took 25 epochs to train and used a dropout rate of 0.15 to avoid overfitting.
  • The model achieved a high coefficient of determination (R = 0.971) and a root mean square error of 3.59 for a 1-hour time lag.
  • The model exhibited consistent accuracy across different seasons.

Sources:

  • VerticalNews. Xi'an University of Technology Reports Findings in Science (Attention mechanism based CNN-LSTM hybrid deep learning model for atmospheric ozone concentration prediction). Chemicals & Chemistry. July 18, 2025; p 4534.
  • Attention mechanism based CNN-LSTM hybrid deep learning model for atmospheric ozone concentration prediction. Scientific Reports, 2025;15(1):21260.

Read more