Air Pollution Forecasting Using Deep Learning Models Based With Hybrid Approach: A Review

Abstract

This study investigates an innovative Deep Learning paradigm for air pollution forecasting, employing a fusion of 1D Convolutional neuronal networks (CNNs) & Bidirectional Gated Recurrent Unit (GRU) architectures. The provided text is insufficient and lacks the necessary details to be rewritten in a clear and concise manner. integration of these models aims to enhance the predictive capabilities by capturing intricate temporal patterns and contextual dependencies within air quality time-series data. CNNs, recognized for their adeptness in discerning temporal features, excel in extracting nuanced patterns from sequential air quality measurements. These networks operate effectively by assimilating diverse data sources, including pollutant concentrations, meteorological variables, and historical trends, facilitating the extraction of localized dependencies within the data. Complementing this, the Bidirectional GRUs contribute a comprehensive understanding by assimilating information from both past and future sequences. This bidirectional approach allows for a holistic comprehension of the temporal dynamics, illuminating the interplay between historical trends and forthcoming forecasts, crucial in forecasting future pollution levels. The combined utilization of CNNs and Bidirectional GRUs presents a robust framework for comprehending the intricacies inherent in air quality data. The CNNs\' proficiency in feature extraction enables the identification of subtle temporal correlations crucial for accurate forecasting in a multi-dimensional data space. Meanwhile, Bidirectional GRUs\' assimilation of future sequences alongside historical data empowers the model to consider anticipatory trends, thereby enhancing the predictive capacity by anticipating impending meteorological changes or variations in emissions. This integrated approach offers a comprehensive understanding of complex temporal relationships within air quality data, enhancing the model\'s predictive accuracy. Notwithstanding these challenges, the integration of 1D CNNs and Bidirectional GRUs presents a promising avenue for advancing air pollution forecasting, promising more accurate predictions, informed environmental management, and improved public health initiatives.

Introduction

I. INTRODUCTION

Air pollution forecasting involves the application of scientific knowledge and technological tools To forecast the precise constitution of atmospheric contamination in a specific geographical area and time in the atmosphere. This procedure is necessary to precisely forecast the situation. The algorithmic estimation of pollutant concentrations can be converted into the air quality index using the same process as actual measurements are used to calculate the index [1]. Various entities, such as government agencies at the state and municipal levels, along with private companies like Airly, Air Visual, Aerostate, Ambee, BreezoMeter, PlumeLabs, and DRAXIS, offer predictions regarding air pollution. These forecasts are produced by private businesses and organizations. Subsequently, these predictions are distributed to local governments and municipalities. Air pollution prediction entails employing mathematical models, Utilizing statistical analysis & machine learning techniques To forecast the concentrations of atmospheric contaminants in a given area specific area during a given timeframe. These projections are crucial for effectively managing public health hazards, making informed policy decisions, and implementing actions to minimize pollution. One approach to predict exposure to air pollution is by employing mathematical models based on the chemistry and physics of the atmosphere. These models consider climatic elements such as temperature, wind speed, and humidity. Additional factors that are considered encompass emissions originating from corporations, natural sources, and other modes of transportation. These models have the capability to forecast the concentrations of pollutants in various geographical areas through simulating the dispersion and chemical reactions of pollutants in the atmosphere. Another approach entails employing statistical methods to analyze historical pollution data and weather trends [1]–[3]. The ability of statistical models to produce forecasts based on comparable weather patterns that will occur in the future is made possible by the identification of connections between past pollution levels and certain weather circumstances. The forecasting of air pollution is another area in which machine learning algorithms play a crucial role. These algorithms analyze vast datasets in order to uncover complicated correlations between various variables and thereby Enhance the precision of forecasts. Continuous data collection on pollution levels through real-time monitoring stations equipped with sensors is crucial for evaluating and improving forecasting models. These websites assess the concentrations of Pollutants include PM2.5 and PM10 parts, which contain nitrogen dioxide (NO2), atmospheric sulfur dioxide (SO2), gaseous carbon monoxide (CO), and ozone (O3). The user's text is incomplete.real-time data is fed into forecasting models, enabling the models to constantly update and enhance their predictions based on the present circumstances[4].

The accuracy and breadth of air pollution forecasting have both been improved as a result of technological advancements such as remote sensing using satellites and drones. Satellite data can provide a more comprehensive spatial perspective by monitoring contaminants across wide geographical areas. This serves as a complementary method to monitoring that conducted on the ground. Nevertheless, numerous challenges persist in the realm of air pollution prediction. These challenges encompass the intricacy of atmospheric processes, the uncertainty in emissions data, and the requirement for high-quality input data in order to make reliable predictions. Furthermore, unanticipated occurrences such as wildfires or industrial accidents can have a substantial impact on air quality, which contributes to the difficulty of precisely forecasting the situation. Refining models through continuous data Efforts to enhance the accuracy of air pollution forecasting involve activities such as data collection, advancing comprehension of atmospheric dynamics, and incorporating information from diverse sources. Within order to develop more robust forecasting systems that are better able to protect public health and the environment, it is essential for scientists, policymakers, and other stakeholders to work together on collaborative projects [5].

II. LITERATURE REVIEW

Xu 2023 et al. Like most spatiotemporal modeling issues, air quality prediction requires specific components to handle the temporal and spatial dependency in complex systems. So far, TSA and RNN models have only been able to mimic time series that do not include any geographical information. The spatial connection of observation sites required earlier work with graph convolution neural networks (GCNs).

By analyzing past data, we may determine The capabilities and interrelationships of a website. Owing to the limitations of human cognitive abilities, inadequate preexisting knowledge is incapable of fully representing the underlying structure.associated to the station or offer better data for reliable prediction. Here, we present a new type of message-passing network called DGN-AEA, which is an adaptive bidirectional dynamic graph that learns edge properties and model parameters. Our approach can benefit from adaptive edge data provided via end-to-end training even in the absence of any background knowledge. Simplifying the problem. In order to help decision-making studies, model by-products can uncover buried structural information between stations [6].

Cities 2023 et al. An urban area is considered a "smart city" if it enhances conventional networks and services through the use of digital technologies, whether for the purpose of public benefit, resource management, or sustained economic growth. More and more people are choosing to live in cities, which is elevating the status of cities and necessitating rapid expansion to accommodate the varied needs of city dwellers. Smart city objectives can be attained through the intelligent application of massive amounts of data collected by The Internet of Things (IoT) is a network that links together tangible devices, vehicles, appliances, and other objects. These objects are outfitted with sensors, software, and network connectivity, enabling them to collect and exchange data.

A method called machine learning has been employed. extensively studied for time-series forecasting, given its diverse applications in fields like healthcare, weather science, retail, and economics. This article will discuss the analysis of IoT time series data from six smart city regions and the prevalent techniques for forecasting multivariate time series using deep neural networks [7].

Gurumoorthy 2023 et al. Particle and pollutant instability and fluctuation makes air quality prediction (AQP) challenging. Poor air quality in cities has been reported in several nations due to increased emissions of PM2.5. In order to reliably forecast air pollution, this study established a regression model based on optimization. To begin, we analyzed real-time PM2.5 readings obtained from a dataset in Beijing spanning the years 2010 to 2014. The real-time dataset encompassed the cities of Cochin, Hyderabad, Chennai, and Bangalore from 2016 to 2022. After performing data normalization using the Min-Max method, the correlation analysis revealed the presence of highly connected variables. The variables encompass Information regarding the direction of the wind, temperatures, dew point, velocity of the wind, and historical PM2.5 data. Subsequently, significant attributes of variables with a high degree of correlation were chosen utilizing reinforced swarm optimization.(RSO). The most effective features were input into a Bi-GRU model to optimize AQP efficiency [8].

Abimannan 2023 et al. Although Pollution monitoring systems are crucial in reducing air pollution, but their development is a challenging endeavor. Combining federated learning with multi-access processing edge (MEC) technologies can significantly improve the accuracy and effectiveness of these systems. The current document delves into the latest findings about air monitoring networks that are backed by federated learning, often known as MEC. Federated learning has numerous uses in air quality monitoring, one of which is the privacy-preserving MEC Our model training method enhances response times and reduces latency. Data privacy, security, and interpretability of models driven by artificial intelligence are significant concerns for real-time air monitoring systems. Modern air monitoring systems employ cutting-edge methods and technologies to effectively address obstacles and provide precise forecasts of air quality [9].

Huang 2023 et al. To improve the precision of short-term wind output prediction, it is beneficial to consider the spatial-temporal correlation among adjacent wind turbines. We present a new 3D generalized recurrent unit model designed specifically for precise short-term prediction of wind turbine output. The 24 turbines neighboring the target turbine are incorporated into the 3D wind power or weather data matrix through the utilization of a convolutional neural network with three dimensions, or generalized recurrent unit encoders, in order to capture their spatial-temporal characteristics. The GRU decoder utilizes its interconnected layers to generate predicted power values for various timeframes. Based on the findings of the SDWPT, our technique exhibits superior prediction accuracy compared to the BPNN, GRU, and 1D CNN-GRU models. The results indicate that the 3D CNN-GRU demonstrates the most efficient configuration. During a 10-minute forecasting period, the validation data set demonstrates an average increase of 10% in Root Mean Square Error (RMSE) and 11% in Mean Absolute Error (MAE), along with a 1% improvement in the R statistic[10].

Table no. 1 Objectives of the research

Table no. 2 Research gap

III. AIR POLLUTION FORECASTING

Industries, automobiles, and other natural and human-made sources contribute significantly to air pollution, which in turn degrades air quality. According to reports from organizations like the European Environment Agency and the As per the World Health Organization (WHO), being subjected to pollutants increases the likelihood of dying at a young age. Air pollution harms the health of humans and other creatures, as well as flora and historical landmarks. Furthermore, respiratory and cardiovascular diseases are closely associated with particle air pollution. When it enters the alveoli, nitrogen dioxide (NO2), which is often produced by burning fossil fuels, causes bronchitis, pneumonia, emphysema, and other respiratory illnesses. When fuels like coal and petroleum are burned, they release Sulfur dioxide (SO2). At high concentrations, it causes respiratory distress and inflammation of the respiratory tract. Sulfur dioxide (SO2) is an extremely reactive gas that is emitted into the atmosphere by multiple sources. These sources encompass fossil fuels such as Coal and petroleum derivatives, power generation facilities, industrial activities such as steel manufacturing and mining, Transportation vehicles, such as locomotives and ships, along with pulp industries, contribute to emissions, which are also sourced from natural sources. from volcanic eruptions.Volcanic eruptions, for example) [21].

Rising populations and economies mean that The global energy consumption is increasing. The global main energy consumption has undergone a substantial 87% surge, primarily driven by a wide array of energy sources. Fossil fuels, such as natural gas, coal, and oil, serve as examples. Renewable energy sources and fossil fuels. fuels, specifically oil and gas, would be able to meet the anticipated rise in energy use over the next few years. At now, almost 62% of the world's energy comes from the most energy-hungry countries. So, it's crucial to simulate their energy usage in order to have a good idea of how much energy the globe will need in the future. Energy consumption is high in cities and other densely populated urban regions. They cause Global greenhouse gas emissions constitute 80% of the total. total, while their energy consumption represents approximately 75% of global usage, despite only taking up 2% of the land. Greenhouse gas emissions have also skyrocketed during the Industrial Revolution, when the reliance There was an increase in the utilization of fossil fuels. Consequently, a majority of individuals are concerned about climate change and global warming, particularly within the past two decades[22]. Extensive Extensive research has been undertaken to analyze the repercussions of climate change on worldwide economies.1990s. International organizations are committed to addressing the adverse impacts of global warming through the implementation of international and legally binding regulations. Carbon dioxide (CO2) is a highly significant greenhouse gas found in Earth's atmosphere. The energy sector is one of the earliest industries that directly combusts fuels, leading to substantial emissions of CO2. Energy-related CO2 emissions contribute approximately 60% of total anthropogenic greenhouse gas emissions. However, the proportion of these emissions varies significantly among countries due to their diverse energy infrastructures. Given its impact on energy generation and favorable environmental outcomes, natural gas (NG) has emerged as a formidable competitor to oil in the global economy. Over the next two decades, there will be a rise in the utilization Regarding fossil fuels, particularly petroleum and natural gas (NG), in order to satisfy the increasing need for energy particularly in the generation of electricity.

A. Techniques Air pollution forecasting

Environmental contamination caused by the presence of harmful substances in the air.forecasting entails the anticipation of pollutant concentrations in the atmosphere during a defined timeframe. Several techniques are used for air pollution forecasting, leveraging various data sources and methodologies:

1. Statistical Methods: These methods utilize historical pollution data and meteorological information to generate models that predict future pollution levels. Regression modeling, data series analysis, especially ARIMA models are employed. are commonly used statistical methodologies.
2. Machine Learning and AI: Algorithms like Artificial intelligence, machines with support vectors, random forests, and deep learning are all types of machine learning algorithms.are all types of machine learning algorithms. models can examine intricate connections among air quality parameters, meteorological data, and other influential factors to predict pollution levels[23].
3. Chemical Transport Models (CTMs): CTMs Model the movement, conversion, and settling of contaminants in the air using mathematical equations based on atmospheric physics and chemistry. These models require extensive computational power and detailed input data, such as emission inventories, meteorological data, and chemical reactions.
4. Hybrid Models: Combining statistical methods with CTMs or machine learning techniques often produces more accurate forecasts. These hybrid models leverage the strengths of different approaches to improve predictions.[24].

B. Challenges of Atmospheric contamination forecasting

Environmental contamination caused by the release of harmful substances into the air.forecasting faces several challenges that can affect the accuracy and reliability of predictions:

1. Complexity of Atmospheric Systems: The atmosphere is a complex system influenced by various factors such as meteorology, geography, emissions from diverse sources, chemical reactions, and atmospheric dynamics. Modeling these interactions accurately is challenging.
2. Data Availability and Quality: Forecasting requires extensive and accurate data on emissions, meteorology, topography, and air quality measurements. Inaccurate or sparse data can lead to less reliable forecasts.
3. Uncertainties in Emission Inventories: Estimating emissions from various sources like industrial facilities, vehicles, and natural sources involves uncertainties due to incomplete data, changes in technology, and variations in human behavior [25].

C. Air quality forecasting models

Several models are used for air quality forecasting, each with its specific focus, methodology, and application. Here are some prominent types of air quality forecasting models:

1. Chemical Transport Models (CTMs): CTMs Model the movement, scattering, and chemical transformations of contaminants in the air. They utilize intricate mathematical formulas rooted in atmospheric physics and chemistry to simulate the actions of pollutants. The models The CMAQ The Community of Practice Multiscale Quality of Air Modeling System (CMAQ) and CAMx (Full Air Quantity Model with Extensions) are two air quality modeling systems. are extensively utilized. Computerized Tomography Machines.
2. Statistical Models: These models utilize statistical methods to analyze Utilize historical air quality info, meteorological variables, emission inventories, along with other pertinent factors in the analysis. forecast pollutant levels. This category includes time series analysis, regression models, and autoregressive models.
3. Machine Learning (ML) Models: Various Machine learning methodologies, including neural networks, support vector machines with neural networks, random forests, as well as deep learning algorithms, are employed for air quality prediction. Machine learning models possess the capability to capture intricate connections between input variables and levels of pollution, thereby enhancing the accuracy of forecasting.
4. Eulerian Models: These models utilize a grid-based representation of the atmosphere and employ mathematical equations to forecast pollutant concentrations at different locations within the grid. The analysis takes into account emissions, meteorology, and chemical reactions at the level of each individual grid cell. Examples of grid-based One of the models available is The WRF-Chem model refers to the Weather Studies & Forecasting model integrated with Chemistry.

IV. DEEP LEARNING

Machine learning involves a neural network consisting of three or more layers, while Deep learning is a specific subset of machine learning. Neural networks strive to mimic the functioning of the human brain in order to gather knowledge from vast quantities of data, although they do not completely replicate its capabilities. Although a neural network with only one layer can make approximations, the addition of hidden layers can greatly enhance and fine-tune the network's performance, resulting in higher accuracy. Many AI-based applications and services rely on deep learning techniques. These applications and services enhance automation by autonomously carrying out analytical and physical tasks without requiring human involvement. Deep learning technology is commonly used in various applications such as digital assistants, voice-activated TV remotes, and credit card fraud detection. Autonomous vehicles exemplify emerging technologies. Deep learning neural networks, or artificial neural networks, aim to replicate the functions of the human brain through the use of data inputs, weights, and bias. To efficiently identify, classify, and express the elements in the data, these components work together harmoniously. Deep neural networks consist of interconnected nodes organized into multiple layers. Successive layers are added to improve the precision and effectiveness of forecasting or classification. Forward propagation is the process of transmitting computations across the network. The visible layers in a deep neural network are the input and output layers, which are readily observable. In a deep learning model, the input layer is responsible for receiving the data that needs to be processed, while the output layer is responsible for generating the final prediction or classification. Back propagation is an algorithmic technique used to compute prediction errors by employing methods like gradient descent. Afterwards, the function's weights and biases are modified by iteratively moving in reverse through the layers, thereby training the model.

A. Deep Learning Model Based on 1D Convents

A Deep Learning model utilizing 1D Convolutional Neural Networks (CNNs) is a powerful approach for sequential data analysis, including time-series data like air quality measurements. Here's an outline of how a 1D CNN model might be structured for air quality forecasting:

1. Data Preparation

a. Input Data: Sequential air quality data containing pollutant concentrations (such as PM2.5, NO2, etc.) along with corresponding timestamps and potentially other relevant features (e.g., meteorological data).

b. Preprocessing: Normalize the data, handle missing values, and potentially resample or aggregate the data to a regular time interval.

c. Model Architecture: A simplified architecture for a 1D CNN model might include the following layers:

2. Training and Optimization

• Loss Function: Specify a suitable The loss operation, such as mean squared error for regression tasks, is Utilized for quantifying the difference between anticipated and observed values.
•  Optimizer: Employ optimization algorithms such as Adam, RMSprop, or SGD to minimize the loss function.
• Training: Train the model using historical air quality data, potentially using a sliding window approach to predict future time steps.

3. Model Evaluation

a. Validation Set: Allocate a distinct subset of the dataset as a validation set to evaluate the effectiveness of the model while it is being trained.

b. Metrics: Assess The performance of the model can be assessed by utilizing evaluation metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE).or R-squared.(for tasks involving the prediction of continuous values)

4. Considerations

a. Hyperparameter Tuning: Experiment with different architectures, kernel sizes, learning rates, and dropout rates to optimize model performance.

b. Overfitting: Employ techniques like dropout, regularization, or early stopping to prevent overfitting.

5. Deployment and Forecasting

a. Forecasting: After undergoing training, the model can be utilized to forecast air quality by providing sequential input data. in subsequent time intervals.

B. Bidirectional GRU

In order to forecast air quality, suitable techniques for Forecasting Time-series duties, such as forecasting air quality, are accessible. The effectiveness of the input data incorporates information from both past and future sequences. It's particularly effective for sequential data analysis, like time-series forecasting, where context from both directions might be crucial.

1. Bidirectional GRU Architecture

• Input Sequences: Sequential input data containing historical air quality measurements typically represented as a time-series.

2. Bidirectional Layer

• Bidirectional GRU layer processes the input sequences in two orientations: anteriorly (from previous to subsequent) and posteriorly (from subsequent to previous).
• Each GRU unit captures temporal dependencies and patterns in the data.

3. Merge Layer

• Combines the outputs from the forward and backward GRU layers. Common methods include concatenation or summation of the outputs.

4. Dense Layers (Optional)

• Fully connected layers might be added on top of the merged outputs for higher-level feature extraction and forecasting.

5. Output Layer

• Final layer for predicting future air quality measurements.

6. Training and Optimization

• Loss Function: In forecasting tasks, it is common to use a regression-based loss function such as the mean square error (MSE) or the mean absolute error (MAE).
• Optimizer: Use optimization techniques like Adam, RMSprop, or SGD to minimize the loss and update model parameters during training.
• Training: Train the bidirectional GRU model using historical air quality data, considering proper data splitting into training, validation, and test sets.

7. Considerations

• Hyper parameter Tuning: Experiment with the number of GRU units, layers, learning rates, and dropout rates to optimize model Efficiency.
• Regularization: Employ Methods such as dropout or L2 regularization can be employed to mitigate the occurrence of undesirable outcomes.over fitting, especially when dealing with limited data.

8. Benefits of Bidirectional GRU

• Context Awareness: It captures information from both past and future sequences, allowing the model to understand broader context and dependencies in the data.
• Improved Learning Representations: By considering bidirectional information flow, it can learn more robust representations of temporal patterns.
• Better Predictive Performance: Bidirectional models often outperform unidirectional ones, especially when future context is important for the prediction task.

9. Deployment and Forecasting

Once trained, the bidirectional GRU model Historical data can be utilized to predict air quality.prediction. The model can generate forecasts for future time steps based on the learned patterns from both directions of the sequence.

Conclusion

Air pollution forecasting is a critical component of environmental management and public health initiatives, necessitating advanced models capable of capturing the complexities inherent in atmospheric dynamics. In this context, the fusion of Deep Learning methodologies, specifically the integration of One-dimensional neural networks using convolution (CNNs) and This relationship Gated Recurrent Units ( GRU ) are used. architectures, emerges as a promising avenue for accurate and comprehensive forecasting. This combined model capitalizes on the distinctive strengths of CNNs and Bidirectional GRUs, addressing the multifaceted nature of air quality data. CNNs, renowned for their adeptness in analyzing temporal patterns, excel in extracting intricate features from sequential air quality measurements. By accommodating multiple input channels encompassing pollutant concentrations, meteorological variables, and historical trends, CNNs offer a robust framework for capturing localized dependencies within the data. Meanwhile, the Bidirectional GRUs contribute a contextual understanding by assimilating Data derived from both historical and prospective sequences. This bidirectional approach allows the model to distinguish more extensive temporal patterns.dependencies, elucidating how historical trends and future forecasts interplay to influence air quality dynamics. The synergy between CNNs and Bidirectional GRUs presents a holistic framework that augments the model\'s capacity to comprehend the nuances within air quality time-series data. The CNNs act as adept feature extractors, identifying subtle temporal patterns and correlations within the multi-dimensional data space. This capability is particularly significant in air quality forecasting, where pollution levels are influenced by a myriad of factors, including emissions, weather conditions, and geographical attributes. Concurrently, the Bidirectional GRUs add a layer of temporal contextuality by assimilating future sequences alongside historical data. This bidirectional insight aids in capturing anticipatory trends, allowing the model to factor in impending meteorological changes or emission variations that might significantly impact future pollution levels. The amalgamation of these architectures encapsulates a comprehensive understanding of the intricate relationships embedded in air quality data, thereby enhancing the model\'s predictive prowess. In conclusion, the integration of a Deep Learning model comprising 1D CNNs and Bidirectional GRUs holds promise in advancing air pollution forecasting capabilities. By synergizing feature extraction and contextual understanding, this model provides a holistic framework to analyze and predict air quality dynamics. Its potential to forecast pollutant concentrations with greater accuracy and comprehend intricate temporal dependencies underscores its significance in bolstering environmental monitoring and fostering informed decision-making for Reducing or lessening The detrimental effects of air pollution have both general health and the environment natural world.

References

[1] S. Yarragunta, M. A. Nabi, P. Jeyanthi, and S. Revathy, “Prediction of air pollutants using supervised machine learning,” Proc. - 5th Int. Conf. Intell. Comput. Control Syst. ICICCS 2021, no. Iciccs, pp. 1633–1640, 2021, doi: 10.1109/ICICCS51141.2021.9432078. [2] G. Isam Drewil and R. Jabbar Al-Bahadili, “Multicultural Education Forecast Air Pollution in Smart City Using Deep Learning Techniques: A Review,” Multicult. Educ., vol. 7, no. 5, pp. 38–47, 2021, doi: 10.5281/zenodo.4737746. [3] D. K. Rajakumari and P. V, “Air Pollution Prediction in Smart Cities by using Machine Learning Techniques,” Int. J. Innov. Technol. Explor. Eng., vol. 9, no. 5, pp. 1272–1279, 2020, doi: 10.35940/ijitee.e2690.039520. [4] Doreswamy, K. S. Harishkumar, Y. Km, and I. Gad, “Forecasting Air Pollution Particulate Matter (PM2.5) Using Machine Learning Regression Models,” Procedia Comput. Sci., vol. 171, no. 2019, pp. 2057–2066, 2020, doi: 10.1016/j.procs.2020.04.221. [5] M. Arsov et al., “Short-term air pollution forecasting based on environmental factors and deep learning models,” Proc. 2020 Fed. Conf. Comput. Sci. Inf. Syst. FedCSIS 2020, vol. 21, pp. 15–22, 2020, doi: 10.15439/2020F211. [6] J. Xu et al., “Dynamic graph neural network with adaptive edge attributes for air quality prediction: A case study in China,” Heliyon, vol. 9, no. 7, p. e17746, 2023, doi: 10.1016/j.heliyon.2023.e17746. [7] S. Cities, “smart cities Multivariate Time-Series Forecasting?: A Review of Deep Learning Methods in Internet of Things Applications to Smart Cities,” pp. 2519–2552, 2023. [8] S. Gurumoorthy, A. K. Kokku, P. Falkowski-Gilski, and P. B. Divakarachari, “Effective Air Quality Prediction Using Reinforced Swarm Optimization and Bi-Directional Gated Recurrent Unit,” Sustain., vol. 15, no. 14, pp. 1–19, 2023, doi: 10.3390/su151411454. [9] S. Abimannan et al., “Towards Federated Learning and Multi-Access Edge Computing for Air Quality Monitoring: Literature Review and Assessment,” Sustainability, vol. 15, no. 18, p. 13951, 2023. [10] X. Huang, Y. Zhang, J. Liu, X. Zhang, and S. Liu, “A Short-Term Wind Power Forecasting Model Based on 3D Convolutional Neural Network – Gated Recurrent Unit,” 2023. [11] A. Shakya and J. S. Rajput, “Prediction of Delhi ’ s Air Quality Index Using Deep Learning Approach,” Int. J. Innov. Sci. Res. Technol., vol. 7, no. 7, pp. 719–725, 2022. [12] V. Barot and V. Kapadia, “Long Short Term Memory Neural Network-Based Model Construction and Fne-Tuning for Air Quality Parameters Prediction,” Cybern. Inf. Technol., vol. 22, no. 1, pp. 171–189, 2022, doi: 10.2478/cait-2022-0011. [13] G. I. Drewil and R. J. Al-Bahadili, “Air pollution prediction using LSTM deep learning and metaheuristics algorithms,” Meas. Sensors, vol. 24, no. September, p. 100546, 2022, doi: 10.1016/j.measen.2022.100546. [14] Q. Zhang, Y. Han, V. O. K. Li, and J. C. K. Lam, “Deep-AIR: A Hybrid CNN-LSTM Framework for Fine-Grained Air Pollution Estimation and Forecast in Metropolitan Cities,” IEEE Access, vol. 10, pp. 55818–55841, 2022, doi: 10.1109/ACCESS.2022.3174853. [15] L. Gao, C. Cai, and X. M. Hu, “Air Quality Prediction Using Machine Learning,” Mach. Learn. Chem. Saf. Heal. Fundam. with Appl., vol. 9, no. 5, pp. 267–288, 2022, doi: 10.1002/9781119817512.ch11. [16] D. Kothandaraman et al., “Intelligent Forecasting of Air Quality and Pollution Prediction Using Machine Learning,” Adsorpt. Sci. Technol., vol. 2022, 2022, doi: 10.1155/2022/5086622. [17] G. Mani, V. Joshi Kumar, and A. A. Stonier, “Prediction and forecasting of air quality index in Chennai using regression and ARIMA time series models,” J. Eng. Res., vol. 10, no. 2 A, pp. 179–194, 2022, doi: 10.36909/jer.10253. [18] S. Sonawani, K. Patil, and P. Chumchu, “NO2 pollutant concentration forecasting for air quality monitoring by using an optimised deep learning bidirectional GRU model,” Int. J. Comput. Sci. Eng., vol. 24, no. 1, pp. 64–73, 2021, doi: 10.1504/IJCSE.2021.113652. [19] S. Singh, A. Yadav, and A. Kumar, “Prediction of Air Pollution Using Random Forest,” vol. 25, no. 9, pp. 1583–6258, 2021, [Online]. Available: http://annalsofrscb.ro [20] I. Mokhtari, W. Bechkit, H. Rivano, and M. R. Yaici, “Uncertainty-Aware Deep Learning Architectures for Highly Dynamic Air Quality Prediction,” IEEE Access, vol. 9, pp. 14765–14778, 2021, doi: 10.1109/ACCESS.2021.3052429. [21] S. Jeya and L. Sankari, “Air Pollution Prediction by Deep Learning Model,” Proc. Int. Conf. Intell. Comput. Control Syst. ICICCS 2020, no. Iciccs, pp. 736–741, 2020, doi: 10.1109/ICICCS48265.2020.9120932. [22] S. Ameer et al., “Comparative Analysis of Machine Learning Techniques for Predicting Air Quality in Smart Cities,” IEEE Access, vol. 7, pp. 128325–128338, 2019, doi: 10.1109/ACCESS.2019.2925082. [23] N. Sharma, S. Taneja, V. Sagar, and A. Bhatt, “Forecasting air pollution load in Delhi using data analysis tools,” Procedia Comput. Sci., vol. 132, pp. 1077–1085, 2018, doi: 10.1016/j.procs.2018.05.023. [24] K. Srinivasa Rao, Dr. G. Lavanya Devi, and N. Ramesh, “Air Quality Prediction in Visakhapatnam using Deep Learning techniques: RNN, LSTM and GRU,” Semant. Sch., vol. 6, no. 3, pp. 2101–2105, 2018, [Online]. Available: http://nebula.wsimg.com/853d3dc4b7598d9a32a2e23f6ba4c1fb?AccessKeyId=DFB1BA3CED7E7997D5B1&alloworigin=1&disposition=0 [25] S. Banara, T. Singh, Y. Thenuia, H. Nandanwar, and A. Chauhan, “Air Pollution Forecasting using Machine Learning and Deep Learning Techniques,” vol. 18, no. 5, pp. 3–7, [Online]. Available: http://xisdxjxsu.asia

Copyright

Copyright © 2024 Sagar Shrivastava, Shyamol Banerjee. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.