Machine-Learning Prediction of Urban Air-Quality Index Using Meteorological and Traffic Variables: Comparison of Random Forest, XGBoost, and LSTM Models
DOI:
https://doi.org/10.66687/JMRISKeywords:
Air quality index, PM2.5, Random ForestAbstract
Background: Urban air quality varies over short time scales because emissions interact with meteorology, traffic intensity, atmospheric chemistry, and pollutant persistence. Data-driven forecasting may improve early warning, but model performance depends on whether algorithms can represent nonlinear tabular relationships or temporal dependencies.
Objective: To compare pre-2023 evidence for Random Forest, XGBoost, and Long Short-Term Memory models in predicting urban air-quality indices or major pollutant concentrations using meteorological, traffic, and historical air-quality variables.
Methods: A structured evidence synthesis was conducted using Q1 journal literature published through 31 December 2022. Studies were included when they evaluated machine-learning or deep-learning prediction of urban AQI, PM2.5, PM10, or related pollution outcomes and reported transferable information concerning meteorological, traffic, temporal, or pollutant predictors. Because datasets, forecast horizons, outcome scales, and validation schemes differed substantially, model metrics were not statistically pooled.
Results: Random Forest performed strongly in tabular and IoT-oriented settings and was particularly useful when traffic and sensor variables were available. XGBoost demonstrated strong nonlinear prediction and bias-correction capability in PM2.5 forecasting, including Shanghai applications integrating meteorology and pollutant information. LSTM and CNN-LSTM approaches showed consistent advantages when historical sequences and multi-hour forecasting were central. Traffic-flow variables improved local roadside prediction, whereas wind speed, temperature, humidity, boundary-layer conditions, and lagged pollutant concentrations were repeatedly influential meteorological or temporal predictors.
Conclusion: No model family is universally superior. Random Forest is a robust default for structured urban sensor data, XGBoost is attractive for high-dimensional nonlinear tabular relationships, and LSTM is best justified when long temporal sequences materially improve the forecast. Fair comparison requires identical temporal splits, features, horizons, and independent test periods.
References
Ameer S, Shah MA, Khan A, Song H, Maple C, Islam SU, Asghar MN. Comparative analysis of machine learning techniques for predicting air quality in smart cities. IEEE Access. 2019;7:128325-128338. doi:10.1109/ACCESS.2019.2925082.
Suleiman A, Tight MR, Quinn AD. Applying machine learning methods in managing urban concentrations of traffic-related particulate matter (PM10 and PM2.5). Atmos Pollut Res. 2019;10(1):134-144. doi:10.1016/j.apr.2018.07.001.
Ma J, Yu Z, Qu Y, Xu J, Cao Y. Application of the XGBoost machine learning method in PM2.5 prediction: a case study of Shanghai. Aerosol Air Qual Res. 2020;20(1):128-138. doi:10.4209/aaqr.2019.08.0408.
Martín-Baos JA, Rodriguez-Benitez L, García-Ródenas R, Liu J. IoT based monitoring of air quality and traffic using regression analysis. Appl Soft Comput. 2022;115:108282. doi:10.1016/j.asoc.2021.108282.
Chang YS, Chiao HT, Abimannan S, Huang YP, Tsai YT, Lin KM. An LSTM-based aggregated model for air pollution forecasting. Atmos Pollut Res. 2020;11(8):1451-1463. doi:10.1016/j.apr.2020.05.015.
Li T, Hua M, Wu X. A hybrid CNN-LSTM model for forecasting particulate matter (PM2.5). IEEE Access. 2020;8:26933-26940. doi:10.1109/ACCESS.2020.2971348.
Yan R, Liao J, Yang J, Sun W, Nong M, Li F. Multi-hour and multi-site air quality index forecasting in Beijing using CNN, LSTM, CNN-LSTM, and spatiotemporal clustering. Expert Syst Appl. 2021;169:114513. doi:10.1016/j.eswa.2020.114513.
Zhang J, Li S. Air quality index forecast in Beijing based on CNN-LSTM multi-model. Chemosphere. 2022;308:136180. doi:10.1016/j.chemosphere.2022.136180.
Wu Q, Lin H. A novel optimal-hybrid model for daily air quality index prediction considering air pollutant factors. Sci Total Environ. 2019;683:808-821. doi:10.1016/j.scitotenv.2019.05.289.
Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput. 1997;9(8):1735-1780. doi:10.1162/neco.1997.9.8.1735.
Sarkar N, Gupta R, Keserwani P. Air Quality Index prediction using an effective hybrid deep learning model. Environ Pollut. 2022;315:120404. doi:10.1016/j.envpol.2022.120404.
Kshirsagar A, Shah M. Anatomization of air quality prediction using neural networks, regression and hybrid models. J Clean Prod. 2022;369:133383. doi:10.1016/j.jclepro.2022.133383.
Kothandaraman D, Praveena N, Varadarajkumar K, Madhav Rao B, Dhanabalan SS, Essa MM, Subramaniam U. Intelligent forecasting of air quality and pollution prediction using machine learning. Adsorpt Sci Technol. 2022;2022:5086622. doi:10.1155/2022/5086622.
Samal KKR, Babu KS, Das SK, Acharaya A. An improved pollution forecasting model with meteorological impact analysis and deep feature extraction. Sustain Cities Soc. 2021;72:103034. doi:10.1016/j.scs.2021.103034.
Cabaneros SM, Calautit JK, Hughes BR. A review of artificial neural network models for ambient air pollution prediction. Environ Model Softw. 2019;119:285-304. doi:10.1016/j.envsoft.2019.06.014.
