Abstract:

Air pollution is among the most critical challenges faced by metropolitan areas, leaving serious impacts on human health and environmental sustainability. Employing accurate and intelligent methods for air quality prediction can play a vital role in managerial decision-making and mitigating the adverse effects of pollutants. In this study, aiming to analyze the factors contributing to the increase of air pollutants in Hamedan city, a hybrid model based on Recurrent Neural Networks (RNN) was developed. Data spanning two and a half consecutive years were collected, preprocessed, and used for model training. The proposed model integrates air pollutant concentrations including PM2.5, PM10, O₃, NO, NOx, NO₂, SO₂, and CO, along with meteorological parameters such as temperature, humidity, wind speed and direction, as well as data related to vehicle traffic and fleet deterioration, to predict pollutant levels for the next 24 hours. The results showed that the proposed model achieved an accuracy of over 97%, demonstrating its high reliability and potential as an effective tool for air quality management and public health protection.

Rapid urbanization, expansion of industrial activities, and increased dependence on motor vehicles in large and industrial cities have led to a significant decline in air quality and an increase in environmental pollution. Air pollution not only threatens human health, but also has devastating effects on ecosystems and the process of climate change. This complex and multifaceted phenomenon encompasses diverse fields, including physics, chemistry, engineering, medicine, and economics, and requires interdisciplinary studies. The lived experience of people in polluted cities, including contracting respiratory diseases or observing human damage caused by pollution, has provided a strong incentive to address this issue. Given the harmful effects of air pollution on human health and other living organisms, this issue has become one of the priorities of public health [1,2]. Numerous studies have shown that there is a direct relationship between air pollution and the prevalence of cardiovascular and respiratory diseases [3]. In response to these challenges, researchers have attempted to provide reliable information to urban managers, planners, and the general public by developing air pollution prediction systems. Given the complex and variable nature of air pollution in both temporal and spatial dimensions, its modeling and prediction are associated with particular difficulties [4]. Therefore, the use of modern methods such as artificial intelligence and neural-fuzzy models has been considered; because these methods have a high ability to analyze complex and nonlinear systems. One of the effective tools in reducing data dimensions and selecting key inputs is principal component analysis (PCA).Rapid urbanization, expansion of industrial activities, and increased dependence on motor vehicles in large and industrial cities have led to a significant decline in air quality and an increase in environmental pollution. Air pollution not only threatens human health, but also has devastating effects on ecosystems and the process of climate change. This complex and multifaceted phenomenon encompasses diverse fields, including physics, chemistry, engineering, medicine, and economics, and requires interdisciplinary studies. The lived experience of people in polluted cities, including contracting respiratory diseases or observing human damage caused by pollution, has provided a strong incentive to address this issue. Given the harmful effects of air pollution on human health and other living organisms, this issue has become one of the priorities of public health [1,2]. Numerous studies have shown that there is a direct relationship between air pollution and the prevalence of cardiovascular and respiratory diseases [3]. In response to these challenges, researchers have attempted to provide reliable information to urban managers, planners, and the general public by developing air pollution prediction systems. Given the complex and variable nature of air pollution in both temporal and spatial dimensions, its modeling and prediction are associated with particular difficulties [4]. Therefore, the use of modern methods such as artificial intelligence and neural-fuzzy models has been considered; because these methods have a high ability to analyze complex and nonlinear systems. One of the effective tools in reducing data dimensions and selecting key inputs is principal component analysis (PCA). This method helps to simplify the modeling process by creating new independent variables [5]. Also, due to the nonlinear and complex behavior of temperature and pollutant concentration changes, traditional modeling methods do not meet the needs of accurate prediction [6]. The use of intelligent models in predicting complex environmental phenomena Given the inherent complexities of phenomena such as air pollution, the use of intelligent methods based on neural networks and neural-fuzzy systems has been considered as efficient tools in modeling and predicting these processes [7]. One of the prominent models in this field is the Adaptive Neuro-Fuzzy Inference System (ANFIS), which, by combining fuzzy logic and the learning ability of neural networks, has a high capability in analyzing and predicting nonlinear and complex behaviors [8]. Numerous studies have shown that evolutionary learning algorithms such as the Particle Swarm Optimization (PSO) algorithm can improve the performance of neural networks and prevent them from getting stuck in local optimum points [5,6]. For example, Cheng et al. (2012) successfully predicted the inflow of the Hongqidi Dam in China by combining an artificial neural network (ANN) and a hybrid PSO algorithm [8].