Yong Zhang , Wentao Cheng , Jie Sun , Li Wang , Shurui Fan , Jingyu Zhang , Shuhao Jiang
{"title":"An adaptive hybrid MEMD-Transformer-BiGRU model using MPA algorithm for air quality prediction","authors":"Yong Zhang , Wentao Cheng , Jie Sun , Li Wang , Shurui Fan , Jingyu Zhang , Shuhao Jiang","doi":"10.1016/j.swevo.2025.102099","DOIUrl":null,"url":null,"abstract":"<div><div>Accurate air quality prediction is usually difficult to achieve because of multiple influencing factors, complex interrelationships and multi-scale processing behaviors. In this study, an adaptive hybrid modelling framework was proposed, which combines Multivariate Empirical Mode Decomposition (MEMD), and Transformer-Bidirectional Gated Recurrent Unit (BiGRU) architectures optimized by the Marine Predators (MPA) Algorithm.</div><div>Firstly, a multi-dimensional feature matrix is constructed considering the multiple input variables such as various pollutants, meteorological factors and air quality indices (AQI). And the MEMD method is used for the matrix decomposition and nonlinear coupling features effective extraction, which could reveal deep coupling relationships and spatiotemporal variation patterns from the multi-source heterogeneous data and the Intrinsic Mode Functions (IMFs) could be obtained.</div><div>Subsequently, a heterogeneous prediction model combining Transformer and BiGRU networks is designed to capture the overall trends and features of the IMFs. In which, the MPA is utilized for the parameters optimization and an adaptive BiGRU network is employed for the weights dynamical adjusting to emphasize the relative importance of each IMF over time.</div><div>Finally, the experiments were given and analyzed with the air quality datasets in Beijing,Tianjin and Shijiazhuang. The results demonstrated that the proposed hybrid model exhibits remarkable efficacy in features extraction, with a root mean square error (RMSE) of 3.1175, mean absolute percentage error (MAPE) of 1.8731 %, and mean absolute error (MAE) of 2.0258, outperforming other comparative models in all comprehensive metrics. Consequently, this hybrid model can effectively improve prediction accuracy and better capture important characteristic information with meteorological factors.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"98 ","pages":"Article 102099"},"PeriodicalIF":8.5000,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Swarm and Evolutionary Computation","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2210650225002573","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0
Abstract
Accurate air quality prediction is usually difficult to achieve because of multiple influencing factors, complex interrelationships and multi-scale processing behaviors. In this study, an adaptive hybrid modelling framework was proposed, which combines Multivariate Empirical Mode Decomposition (MEMD), and Transformer-Bidirectional Gated Recurrent Unit (BiGRU) architectures optimized by the Marine Predators (MPA) Algorithm.
Firstly, a multi-dimensional feature matrix is constructed considering the multiple input variables such as various pollutants, meteorological factors and air quality indices (AQI). And the MEMD method is used for the matrix decomposition and nonlinear coupling features effective extraction, which could reveal deep coupling relationships and spatiotemporal variation patterns from the multi-source heterogeneous data and the Intrinsic Mode Functions (IMFs) could be obtained.
Subsequently, a heterogeneous prediction model combining Transformer and BiGRU networks is designed to capture the overall trends and features of the IMFs. In which, the MPA is utilized for the parameters optimization and an adaptive BiGRU network is employed for the weights dynamical adjusting to emphasize the relative importance of each IMF over time.
Finally, the experiments were given and analyzed with the air quality datasets in Beijing,Tianjin and Shijiazhuang. The results demonstrated that the proposed hybrid model exhibits remarkable efficacy in features extraction, with a root mean square error (RMSE) of 3.1175, mean absolute percentage error (MAPE) of 1.8731 %, and mean absolute error (MAE) of 2.0258, outperforming other comparative models in all comprehensive metrics. Consequently, this hybrid model can effectively improve prediction accuracy and better capture important characteristic information with meteorological factors.
期刊介绍:
Swarm and Evolutionary Computation is a pioneering peer-reviewed journal focused on the latest research and advancements in nature-inspired intelligent computation using swarm and evolutionary algorithms. It covers theoretical, experimental, and practical aspects of these paradigms and their hybrids, promoting interdisciplinary research. The journal prioritizes the publication of high-quality, original articles that push the boundaries of evolutionary computation and swarm intelligence. Additionally, it welcomes survey papers on current topics and novel applications. Topics of interest include but are not limited to: Genetic Algorithms, and Genetic Programming, Evolution Strategies, and Evolutionary Programming, Differential Evolution, Artificial Immune Systems, Particle Swarms, Ant Colony, Bacterial Foraging, Artificial Bees, Fireflies Algorithm, Harmony Search, Artificial Life, Digital Organisms, Estimation of Distribution Algorithms, Stochastic Diffusion Search, Quantum Computing, Nano Computing, Membrane Computing, Human-centric Computing, Hybridization of Algorithms, Memetic Computing, Autonomic Computing, Self-organizing systems, Combinatorial, Discrete, Binary, Constrained, Multi-objective, Multi-modal, Dynamic, and Large-scale Optimization.