Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization

To address the performance degradation in existing PM<sub>2.5</sub> prediction models caused by excessive complexity, poor spatiotemporal efficiency, and suboptimal parameter optimization, we employ stacking ensemble learning for feature weighting analysis and integrate the ant colony op...

Full description

Saved in:
Bibliographic Details
Main Authors: Zuhan Liu, Xianping Hong
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Toxics
Subjects:
Online Access:https://www.mdpi.com/2305-6304/13/5/327
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849718915575316480
author Zuhan Liu
Xianping Hong
author_facet Zuhan Liu
Xianping Hong
author_sort Zuhan Liu
collection DOAJ
description To address the performance degradation in existing PM<sub>2.5</sub> prediction models caused by excessive complexity, poor spatiotemporal efficiency, and suboptimal parameter optimization, we employ stacking ensemble learning for feature weighting analysis and integrate the ant colony optimization (ACO) algorithm for model parameter optimization. Combining meteorological and collaborative pollutant data, a model (namely the stacking-ACO-LSTM model) with a much shorter consuming time than that of only long short-term memory (LSTM) networks suitable for PM<sub>2.5</sub> concentration prediction is established. It can effectively filter out feature variables with higher weights, thereby reducing the predictive power of the model. The prediction of hourly PM<sub>2.5</sub> concentration of the model is trained and tested using real-time monitoring data in Nanchang City from 2017 to 2019. The results show that the established stacking-ACO-LSTM model has high accuracy in predicting PM<sub>2.5</sub> concentration, and compared to the same model without considering time and space efficiency and defective parameter optimization, the mean square error (MSE) decreases by about 99.88%, and the coefficient of determination (R<sup>2</sup>) increases by about 2.39%. This study provides a new idea for predicting PM<sub>2.5</sub> concentration in cities.
format Article
id doaj-art-d74c15b460dd4d278bb68878464b0ac2
institution DOAJ
issn 2305-6304
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Toxics
spelling doaj-art-d74c15b460dd4d278bb68878464b0ac22025-08-20T03:12:15ZengMDPI AGToxics2305-63042025-04-0113532710.3390/toxics13050327Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony OptimizationZuhan Liu0Xianping Hong1School of Information Engineering, Nanchang Institute of Technology, Nanchang 330099, ChinaSchool of Information Engineering, Nanchang Institute of Technology, Nanchang 330099, ChinaTo address the performance degradation in existing PM<sub>2.5</sub> prediction models caused by excessive complexity, poor spatiotemporal efficiency, and suboptimal parameter optimization, we employ stacking ensemble learning for feature weighting analysis and integrate the ant colony optimization (ACO) algorithm for model parameter optimization. Combining meteorological and collaborative pollutant data, a model (namely the stacking-ACO-LSTM model) with a much shorter consuming time than that of only long short-term memory (LSTM) networks suitable for PM<sub>2.5</sub> concentration prediction is established. It can effectively filter out feature variables with higher weights, thereby reducing the predictive power of the model. The prediction of hourly PM<sub>2.5</sub> concentration of the model is trained and tested using real-time monitoring data in Nanchang City from 2017 to 2019. The results show that the established stacking-ACO-LSTM model has high accuracy in predicting PM<sub>2.5</sub> concentration, and compared to the same model without considering time and space efficiency and defective parameter optimization, the mean square error (MSE) decreases by about 99.88%, and the coefficient of determination (R<sup>2</sup>) increases by about 2.39%. This study provides a new idea for predicting PM<sub>2.5</sub> concentration in cities.https://www.mdpi.com/2305-6304/13/5/327PM<sub>2.5</sub>long short-term memory network (LSTM)ant colony optimization (ACO)stacking ensemble learningdeep learning
spellingShingle Zuhan Liu
Xianping Hong
Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization
Toxics
PM<sub>2.5</sub>
long short-term memory network (LSTM)
ant colony optimization (ACO)
stacking ensemble learning
deep learning
title Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization
title_full Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization
title_fullStr Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization
title_full_unstemmed Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization
title_short Improved Prediction of Hourly PM<sub>2.5</sub> Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization
title_sort improved prediction of hourly pm sub 2 5 sub concentrations with a long short term memory optimized by stacking ensemble learning and ant colony optimization
topic PM<sub>2.5</sub>
long short-term memory network (LSTM)
ant colony optimization (ACO)
stacking ensemble learning
deep learning
url https://www.mdpi.com/2305-6304/13/5/327
work_keys_str_mv AT zuhanliu improvedpredictionofhourlypmsub25subconcentrationswithalongshorttermmemoryoptimizedbystackingensemblelearningandantcolonyoptimization
AT xianpinghong improvedpredictionofhourlypmsub25subconcentrationswithalongshorttermmemoryoptimizedbystackingensemblelearningandantcolonyoptimization