Article Details

Abstract

The Air Quality Index (AQI), as a key indicator for assessing environmental health, plays a crucial role in environmental policy formulation and public health. However, traditional prediction methods face limitations in handling the complexity of air quality data. To address this limitation, an AQI prediction model based on the Long Short-Term Memory (LSTM) network is proposed. Utilizing nearly a decade of air quality and meteorological data from Nanjing, the model achieves temporal forecasting through feature selection, hyperparameter optimization, and model construction. The results show that the optimized LSTM model achieves a Root Mean Square Error (RMSE) of 21.034 9, a Mean Absolute Percentage Error (MAPE) of 28.10%, and a Coefficient of Determination (R2) of 0.539 0 on the test set, demonstrating higher prediction accuracy than the Support Vector Machine (SVM) model and the BP neural network model. This model provides an effective deep learning solution for dynamic air quality forecasting in urban areas and serves as a technical reference for environmental management decision-making.

Keywords

AQI; LSTM; Time Series Forecasting; hyperparameter optimization; environmental management

References

1. [1] Shi Xueliang, Li Liang, Zhao Qinghua Prediction of Air Quality Index Based on Improved LSTM Network [J]. Statistics and Decision, 2021, 37 (16): 57-60
2. [2] Mou Jingfeng, Zhao Xing, Fan Jingjie, etc Research on Time Series Prediction of Air Quality Index in Shenzhen City Based on ARIMA Model [J]. Journal of Environmental Health, 2017, 7 (2): 102-107117
3. [3] Xiao Delin, Deng Shihuai, Deng Xiaohan, etc Trend of Environmental Air Quality Changes in the Urban Area of Dazhou City and CMAQ Model Forecast Analysis [J]. China Environmental Monitoring, 2021, 37 (4): 92-103
4. [4] Hua Junwei, Bai Lin, Xing Cheng, etc Research on Air Quality Forecasting in Chuzhou City Based on Random Forest Algorithm [J]. Environmental Monitoring Management and Technology, 2024, 36 (1): 74-78
5. [5] Chen Jinche, Dilinuer Yasheng, Wang Tianyu, et al. Comparison of the effectiveness of three machine learning methods in air pollution forecasting [J]. Journal of Lanzhou University (Natural Science Edition), 2023, 59 (5): 638-646
6. [6] Liu Ying, Chen Xudong, Zhou Mi, etc Research on Regional AQI Prediction Based on Multi dimensional Multi step LSTM Network [J]. Industrial Safety and Environmental Protection, 2022, 48 (10): 100-103
7. [7] Ji Danjun Time Series Analysis and Prediction of Urban Air Quality Index: A Case Study of Yangzhou City [J]. Journal of Wuxi Vocational and Technical College of Commerce, 2024, 24 (1): 59-65
8. [8] GRAVES A. Supervised Sequence Labelling with Recurrent Neural Networks [M].Berlin：Springer，2012.
9. [9] Li L S，Kevin J，Giulia D，et al. Hyperband： A Novel Bandit-Based Approach to Hyperparameter Optimization [J].JOURNAL OF MACHINE LEARNING RESEARCH，2018，18（185）：1-52.
10. [10] Jiang Yu, Ma Tinghuai Prediction of winter wheat yield in Henan Province based on CNN-LSTM Attention network [J]. Journal of Wheat Crops, 2024, 44 (10): 1352-1359

About the authors

Shi Jiayi (2001-), female, Han ethnicity, from Taizhou, Jiangsu Province, is currently pursuing a master's degree in Operations Research Optimization.