Enhanced Air Quality Prediction Using a Two-Stage XGBoost and LSTM Hybrid Architecture

View at publisher Summarise with VIRAT
Authors: R Rinisri, Ramya K, M. Revathi, Jennifer Mary A.
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The issue of air pollution forecasting has gained urgency due to the accelerated urbanization, industrial processes and climatic nature of fluctuation in the behavior of pollutants. While traditional machine learning and deep learning approaches demonstrate promising results, their major limitations still lie in either capturing nonlinear feature relationships or long-term temporal dependencies. This work presents an integrated XGBoost-LSTM hybrid for air quality prediction, combining gradient boosting for learning feature interactions with recurrent neural networks for temporal sequence modeling. The approach encompasses data preprocessing, feature engineering, residual learning, and multi-stage prediction. Experimental results show the supremacy of the hybrid model over the respective single models by reducing the RMSE to$6.48 \mu ~\mathrm{g} / \mathrm{m}^{3}$, MAE to 4.82$\mu \mathrm{g} / \mathrm{m}^{3}$, and an$\mathrm{R}^{2}$accuracy of 0.96. The system also maintains stable performance over multiple days and can thus be considered for real-time air quality monitoring and decision-support systems.

Fields of Study & Subjects

Air Quality Monitoring and Forecasting Air Quality and Health Impacts Hydrological Forecasting Using AI

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1688–1693
Publication date
21 January 2026
Language
EN