ArticlePDF Available

Deep learning architecture for air quality predictions

Authors:
Xiang Li at King Abdullah University of Science and Technology
Xiang Li
Ling Peng
Ling Peng
Ling Peng
  • This person is not on ResearchGate, or hasn't claimed this research yet.
Yuan Hu
Yuan Hu
Yuan Hu
  • This person is not on ResearchGate, or hasn't claimed this research yet.
Jing Shao at Chinese Academy of Sciences
Jing Shao
Show all 5 authorsHide
Download full-text PDF
Read full-text
Download citation

Abstract and Figures

With the rapid development of urbanization and industrialization, many developing countries are suffering from heavy air pollution. Governments and citizens have expressed increasing concern regarding air pollution because it affects human health and sustainable development worldwide. Current air quality prediction methods mainly use shallow models; however, these methods produce unsatisfactory results, which inspired us to investigate methods of predicting air quality based on deep architecture models. In this paper, a novel spatiotemporal deep learning (STDL)-based air quality prediction method that inherently considers spatial and temporal correlations is proposed. A stacked autoencoder (SAE) model is used to extract inherent air quality features, and it is trained in a greedy layer-wise manner. Compared with traditional time series prediction models, our model can predict the air quality of all stations simultaneously and shows the temporal stability in all seasons. Moreover, a comparison with the spatiotemporal artificial neural network (STANN), auto regression moving average (ARMA), and support vector regression (SVR) models demonstrates that the proposed method of performing air quality predictions has a superior performance.
Figures - available from: Environmental Science and Pollution Research
This content is subject to copyright. Terms and conditions apply.
Content uploaded by Xiang Li
Author content
All content in this area was uploaded by Xiang Li on Dec 19, 2017
Content may be subject to copyright.
A preview of this full-text is provided by Springer Nature.
Learn more
Content available from Environmental Science and Pollution Research
This content is subject to copyright. Terms and conditions apply.
... Data mining allows to analysis of air quality using analytical methods when scientific methods do not exist; in that sense, Soh et al. propose a predictive system for air quality using ST-DNN (Shape-Tailored Deep Neural Networks) to predict PM2.5 48 hours in advance (Soh et al. 2016). Also, Li et al. (2016) applied to predict air quality deep learning approach, applying a regional data treatment as a spatiotemporal process that considers spatial and temporal correlations of data to predict the air quality of all monitoring stations simultaneously with seasonal stability (Li et al. 2016), which recognizes and applies seasonal behavior in analysis and predictions. Wang et al. (2017) categorized the main pollutant forecasting models as deterministic, statistical, and hybrid models (Wang et al. 2017). ...
... Data mining allows to analysis of air quality using analytical methods when scientific methods do not exist; in that sense, Soh et al. propose a predictive system for air quality using ST-DNN (Shape-Tailored Deep Neural Networks) to predict PM2.5 48 hours in advance (Soh et al. 2016). Also, Li et al. (2016) applied to predict air quality deep learning approach, applying a regional data treatment as a spatiotemporal process that considers spatial and temporal correlations of data to predict the air quality of all monitoring stations simultaneously with seasonal stability (Li et al. 2016), which recognizes and applies seasonal behavior in analysis and predictions. Wang et al. (2017) categorized the main pollutant forecasting models as deterministic, statistical, and hybrid models (Wang et al. 2017). ...
Application of Random Forest in a Predictive Model of PM10 Particles in Mexico City
Article
  • Jun 2024
Over time, predictive models tend to become more accurate but also more complex, thus achieving better predictive accuracy. When the data is improved by increasing its quantity and availability, the models are also better, which implies that the data must be processed to filter and adapt it for initial analysis and then modeling. This work aims to apply the Random Forest model to predict PM10 particles. For this purpose, data were obtained from environmental monitoring stations in Mexico City, which operates 29 stations of which 12 belong to the State of Mexico. The pollutants analyzed were CO carbon monoxide, NO nitrogen oxide, and PM10 particulate matter equal to or less than 10 μg.m-3, NOx nitrogen oxide, NO2 nitrogen dioxide, SO2 sulfur dioxide, O3 ozone, and PM2.5 particulate matter equal to or less than 2.5 μg.m-3. The result was that when calculating the certainty of our model, we have a value of 80.40% when calculating the deviation from the mean, using 15 reference variables.
View
... Future research may aim to look for models that fit specific data.. Such research may lead to the use of more complex modeling tools, which have been described in many publications [51][52][53]65]. Research can also be conducted on segment modeling. ...
Regression Modeling of Daily PM2.5 Concentrations with a Multilayer Perceptron
Article
Full-text available
  • May 2024
Various types of energetic fuel combustion processes emit dangerous pollutants into the air, including aerosol particles, marked as PM10. Routine air quality monitoring includes determining the PM10 concentration as one of the basic measurements. At some air monitoring stations, the PM10 measurement is supplemented by the simultaneous determination of the concentration of PM2.5 as a finer fraction of suspended particles. Since the PM2.5 fraction has a significant share in the PM10 fraction, the concentrations of both types of particles should be strongly correlated, and the concentrations of one of these fractions can be used to model the concentrations of the other fraction. The aim of the study was to assess the error of predicting PM2.5 concentration using PM10 concentration as the main predictor. The analyzed daily concentrations were measured at 11 different monitoring stations in Poland and covered the period 2010–2021. MLP (multilayer perceptron) artificial neural networks were used to approximate the daily PM2.5 concentrations. PM10 concentrations and time variables were tested as predictors in neural networks. Several different prediction errors were taken as measures of modeling quality. Depending on the monitoring station, in models with one PM10 predictor, the RMSE error values were in the range of 2.31–6.86 μg/m3. After taking into account the second predictor D (date), the corresponding RMSE errors were lower and were in the range of 2.06–5.54 μg/m3. Our research aimed to find models that were as simple and universal as possible. In our models, the main predictor is the PM10 concentration; therefore, the only condition to be met is monitoring the measurement of PM10 concentrations. We showed that models trained at other air monitoring stations, so-called foreign models, can be successfully used to approximate PM2.5 concentrations at another station.
View
... Air pollution is a serious environmental problem that attracts worldwide attention due to its adverse effects on human health and sustainable development [1]. With the rapid development of industrialization, urbanization, and the economy, many developing countries suffer from heavy air pollution [2]. The main sources of air pollution are particulate matters (PM 2.5 and PM 10 ) and pollutant gases (O 3 , NO 2 , SO 2 , CO) [3]. ...
Cepair: an AI-powered and fog-based predictive CEP system for air quality monitoring
Article
Full-text available
  • Apr 2024
  • CLUSTER COMPUT
Air pollution is one of the influential problems threatening the environment and human health today. Therefore, it is critical to develop predictive systems for proactive decisions in solving this problem. Since the prediction of air pollution depends on several complicated factors such as the accuracy of meteorology reports, air pollution accumulation, traffic flow, and industrial emissions, the contribution of historical or real-time predictions to the solution of the problem is limited. To address the existing limitations, we propose a novel AI-powered and Fog-based predictive complex event processing system (CepAIr) for the prediction of future air pollution rates. CepAIr predicts the future air quality of pollutant gases using RNN, LSTM, CNN, and SVR models. Then, it sends the prediction results to decision-makers in an understandable format, enabling them to take proactive actions. Finally, we evaluate the performance of the CepAIr with SVR and DL models. Additionally, we examine CepAIr in terms of end-to-end network delay and measure its impact on the network. The extensive simulation results demonstrate that the CepAIr predicts future pollutant gas concentrations with DL models (especially with CNN) with a high success rate while guaranteeing minimum end-to-end network delay.
View
... Krishan et al. [47] amalgamated five combinations of factors and parameters, including vehicle emissions, meteorological conditions, traffic data, and pollutant levels, utilizing a two-layer LSTM model to adeptly capture long-term dependencies in modeling air quality parameters. Li et al. [48] proposed a spatiotemporal deep learning (STDL) model that leverages a stacked self-encoder (SAE) model to extract intrinsic features related to air quality. Yan et al. [49] devised a CNN-LSTM prediction model for air quality in Beijing, which incorporates multi-temporal and multisite data. ...
Enhancing Air Quality Forecasting: A Novel Spatio-Temporal Model Integrating Graph Convolution and Multi-Head Attention Mechanism
Article
Full-text available
  • Mar 2024
Forecasting air quality plays a crucial role in preventing and controlling air pollution. It is particularly significant for improving preparedness for heavily polluted weather conditions and ensuring the health and safety of the population. In this study, a novel deep learning model for predicting air quality spatio-temporal variations is introduced. The model, named graph long short-term memory with multi-head attention (GLSTMMA), is designed to capture the temporal patterns and spatial relationships within multivariate time series data related to air quality. The GLSTMMA model utilizes a hybrid neural network architecture to effectively learn the complex dependencies and correlations present in the data. The extraction of spatial features related to air quality involves the utilization of a graph convolutional network (GCN) to collect air quality data based on the geographical distribution of monitoring sites. The resulting graph structure is imported into a long short-term memory (LSTM) network to establish a Graph LSTM unit, facilitating the extraction of temporal dependencies in air quality. Leveraging a Graph LSTM unit, an encoder-multiple-attention decoder framework is formulated to enable a more profound and efficient exploration of spatio-temporal correlation features within air quality time series data. The research utilizes the 2019–2021 multi-source air quality dataset of Qinghai Province for experimental assessment. The results indicate that the model effectively leverages the impact of multi-source data, resulting in optimal accuracy in predicting six air pollutants.
View
... Regarding spatiotemporal methodologies, the prevalent approach combines LSTM for temporal aspects with Convolutional Neural Networks (CNN) for spatial domains, aiming to analyze both temporal and spatial correlations in pollutant concentrations (Bekkar et al., 2021;Mao et al., 2021). However, alternative strategies exist, such as stacked autoencoder models extracting spatiotemporal features in a layer-wise manner (Li et al., 2016), or hierarchical Bayesian spatiotemporal approaches (Saez and Barceló, 2022). Zhang et al. (2022) deep learning model based on the Transformer architecture to learn long-term temporal dependencies for air quality PM2.5 prediction. ...
Spatial air quality prediction in urban areas via message passing
Article
Full-text available
  • Mar 2024
  • ENG APPL ARTIF INTEL
Air pollution in urban areas poses a significant and pressing challenge for modern society. Unfortunately, the existing network of pollution detectors in many cities is limited in scope and fails to adequately cover the entire geographical area. Consequently, the implementation of spatial prediction algorithms becomes essential to generate high-resolution data. In this paper, we introduce two significant contributions: 1) We formalize the air pollution prediction problem as a Maximum A Posteriori (MAP) estimate within the framework of a Markov Random Field and 2) we propose a message-passing algorithm, which stands out as an efficient solution that surpasses the current state of the art. The experimental procedure has been carried out using the case study of the city of Barcelona, based on a dataset extracted from the BCN Open Data portal.
View
... They are used to analyze historical O 3 monitoring data and meteorological conditions, as well as to establish mathematical relationships for predicting future O 3 concentrations, since meteorological factors such as temperature, relative humidity, and solar radiation exhibit a strong correlation with O 3 concentration [18]. The advantage of statistical models is the fast execution speed, but they lack detailed a physical-chemical interpretation of results and are considered black box models [19,20]. O 3 ensemble models [21] utilize statistical methods to analyze and compute results based on numerical model forecasts combined with observed data, ultimately providing adjusted O 3 forecast results. ...
Comparative Study of O3 Forecast Performance Using Multiple Models in Beijing–Tianjin–Hebei and Surrounding Regions
Article
Full-text available
  • Feb 2024
In order to systematically understand the operational forecast performance of current numerical, statistical, and ensemble models for O3 in Beijing–Tianjin–Hebei and surrounding regions, a comprehensive evaluation was conducted for the 30 model sets regarding O3 forecasts in June–July 2023. The evaluation parameters for O3 forecasts in the next 1–3 days were found to be more reasonable and practically meaningful than those for longer lead times. When the daily maximum 8 h average concentration of O3 was below 100 μg/m3 or above 200 μg/m3, a significant decrease in the percentage of accurate models was observed. As the number of polluted days in cities increased, the overall percentage of accurate models exhibited a decreasing trend. Statistical models demonstrated better overall performance in terms of metrics such as root mean square error, standard mean bias, and correlation coefficient compared to numerical and ensemble models. Numerical models exhibited significant performance variations, with the best-performing numerical model reaching a level comparable to that of statistical models. This finding suggests that the continuous tuning of operational numerical models has a more pronounced practical effect. Although the best statistical model had higher accuracy than numerical and ensemble models, it showed a significant overestimation when O3 concentrations were low and a significant underestimation when concentrations were high. In particular, the underestimation rate for heavy polluted days was significantly higher than that for numerical and ensemble models. This implies that statistical models may be more prone to missing high-concentration O3 pollution events.
View
View
View
Technological Progression associated with monitoring and management of indoor air pollution and associated health risks: A Comprehensive Review
Article
  • Apr 2024
  • Environ Qual Manag
Indoor air quality (IAQ) is significantly impacted by ambient air pollution, and assessing and understanding pollution causes is crucial for developing mitigation plans. The prospect of a "smart environment" using IoT is being explored to improve indoor living conditions. Technological advancements have simplified the design of ambient assisted living systems. This review explores how IoT can regulate indoor air quality, focusing on models of dispersion, land use, sensors, and their coupling with cellphones, as well as air pollution simulations. The review discusses smart ventilation techniques like RIVEC and exhaust-only DCV systems, as well as computational tools for accurate health risk assessment. It covers challenges like data coding and safety concerns, and future prospects like deep learning models and mobile app-based systems for predictions. The report also includes literature indexed using top databases, concentrating on replicating actual circumstances and real-time IAQ monitoring devices. Keywords: Internet of Things, technological advancements, smart ventilation techniques, Indoor air quality, deep learning models.
View
Comparative Analysis of Machine Learning Techniques in Air Quality Index (AQI) prediction in smart cities
Article
  • Apr 2024
Air pollution is now one of the world's most serious environmental problems. It represents a significant hazard to both health and the climate. Urban air quality is steadily declining, affecting not only the air itself but also impacting the quality of water and land. This paper explores the utilization of machine learning-based algorithms for analysis and prediction of air quality in smart cities. In this paper, smart cities for which air quality index (AQI) is calculated are Ahmedabad, Delhi, Lucknow, Gurugram, and Mumbai. The comparative analysis of different Machine Learning algorithms such as Random Forest Regression (RF), Decision Tree Regression, Linear regression, XgBoost and proposed hybrid model which is combination of Random forest and Xgboost model, have been discussed in the paper. The analysis has been carried out using a machine learning-based algorithm to determine which pollutant is the primary source of pollution in a smart city so that preventative steps can be implemented to reduce air pollution.
View
View
Mastering the game of Go with deep neural networks and tree search
Article
Full-text available
  • Jan 2016
The game of Go has long been viewed as the most challenging of classic games for artificial intelligence owing to its enormous search space and the difficulty of evaluating board positions and moves. Here we introduce a new approach to computer Go that uses ‘value networks’ to evaluate board positions and ‘policy networks’ to select moves. These deep neural networks are trained by a novel combination of supervised learning from human expert games, and reinforcement learning from games of self-play. Without any lookahead search, the neural networks play Go at the level of state-of-the-art Monte Carlo tree search programs that simulate thousands of random games of self-play. We also introduce a new search algorithm that combines Monte Carlo simulation with value and policy networks. Using this search algorithm, our program AlphaGo achieved a 99.8% winning rate against other Go programs, and defeated the human European Go champion by 5 games to 0. This is the first time that a computer program has defeated a human professional player in the full-sized game of Go, a feat previously thought to be at least a decade away.
View
Dynamic Analysis of a Lü Model in Six Dimensions and Its Projections
Article
  • Jul 2017
In this article, extended complex Lü models (ECLMs) are proposed. They are obtained by substituting the real variables of the classical Lü model by complex variables. These projections, spanning from five dimensions (5D) and six dimensions (6D), are studied in their dynamics, which include phase spaces, calculations of eigenvalues and Lyapunov’s exponents, Poincaré maps, bifurcation diagrams, and related analyses. It is shown that in the case of a 5D extension, we have obtained chaotic trajectories; meanwhile the 6D extension shows quasiperiodic and hyperchaotic behaviors and it exhibits strange nonchaotic attractor (SNA) features.
View
View
Temporal trends of spatial correlation within the PM10 time series of the AirBase ambient air quality database
Article
  • Jan 2015
We analyse the temporal variations which can be observed within time series of variogram parameters (nugget, sill and range) of daily air quality data (PM10) over a ten years time frame. Datasets have been obtained from previous geostatistical analysis of country wide datasets from the AirBase ambient air quality database. Applying the Kolmogorov-Zurbenko filtering method, the time series are first decomposed into their short-, mid-, and long-term components. Based on this, we then evaluate the magnitude of the individual spectral signal contributions. Furthermore, the significance of a long term trend component is investigated by a block-bootstrap-based approach combined with linear regression. It is discussed if within these datasets the times series of nugget variance can provide information about the evolution of the measurement uncertainty of the related air pollutant, whereas the sill and the range parameters could contain information about the spatial representativeness of the monitoring stations.
View
Salient Band Selection for Hyperspectral Image Classification via Manifold Ranking
Article
  • Mar 2016
Saliency detection has been a hot topic in recent years, and many efforts have been devoted in this area. Unfortunately, the results of saliency detection can hardly be utilized in general applications. The primary reason, we think, is unspecific definition of salient objects, which makes that the previously published methods cannot extend to practical applications. To solve this problem, we claim that saliency should be defined in a context and the salient band selection in hyperspectral image (HSI) is introduced as an example. Unfortunately, the traditional salient band selection methods suffer from the problem of inappropriate measurement of band difference. To tackle this problem, we propose to eliminate the drawbacks of traditional salient band selection methods by manifold ranking. It puts the band vectors in the more accurate manifold space and treats the saliency problem from a novel ranking perspective, which is considered to be the main contributions of this paper. To justify the effectiveness of the proposed method, experiments are conducted on three HSIs, and our method is compared with the six existing competitors. Results show that the proposed method is very effective and can achieve the best performance among the competitors.
View
View
Predictive Deep Boltzmann Machine for Multi-Period Wind Speed Forecasting
Article
  • Oct 2015
It is important to forecast the wind speed for managing operations in wind power plants. However, wind speed prediction is extremely complex and difficult due to the volatility and deviation of the wind. As existing forecasting methods directly model the raw wind speed data, it is difficult for them to provide higher inference accuracy. Differently, this paper presents a sophisticated deep-learning technique for short-term and long-term wind speed forecast, i.e., the predictive deep Boltzmann machine (PDBM) and corresponding learning algorithm. The proposed deep model forecasts wind speed by analyzing the higher level features abstracted from lower level features of the wind speed data. These automatically learnt features are very informative and appropriate for the prediction. The proposed PDBM is a deep stochastic model that can represent the wind speed very well, and is inspired by two aspects. 1)The stochastic model is suitable to capture the probabilistic characteristics of wind speed. 2)Recent developments in neural networks with deep architectures show that deep generative models have competitive capability to approximate nonlinear and nonsmooth functions. The evaluation of the proposed PDBM model is depicted by both hour-ahead and day-ahead prediction experiments based on real wind speed datasets. The prediction accuracy of the PDBM model outperforms existing methods by more than 10%.
View
Traffic Flow Prediction With Big Data: A Deep Learning Approach
Article
  • Jan 2014
Accurate and timely traffic flow information is important for the successful deployment of intelligent transportation systems. Over the last few years, traffic data have been exploding, and we have truly entered the era of big data for transportation. Existing traffic flow prediction methods mainly use shallow traffic prediction models and are still unsatisfying for many real-world applications. This situation inspires us to rethink the traffic flow prediction problem based on deep architecture models with big traffic data. In this paper, a novel deep-learning-based traffic flow prediction method is proposed, which considers the spatial and temporal correlations inherently. A stacked autoencoder model is used to learn generic traffic flow features, and it is trained in a greedy layerwise fashion. To the best of our knowledge, this is the first time that a deep architecture model is applied using autoencoders as building blocks to represent traffic flow features for prediction. Moreover, experiments demonstrate that the proposed method for traffic flow prediction has superior performance.
View
Nonlinear Air Quality Modeling Using Support Vector Machines in Gijón Urban Area (Northern Spain) at Local Scale
Article
  • Jan 2013
  • INT J NONLIN SCI NUM
Due to the health impacts caused by exposures to air pollutants in urban areas, monitoring and forecasting of air quality parameters have become popular as an important topic in atmospheric and environmental research today. The knowledge on the dynamics and complexity of air pollutants behaviour has made artificial intelligence and statistical learning models as a useful tool for a more accurate pollutant concentration prediction. The aim of this study is to build a regression model of daily air quality prediction by using support vector machine (SVM) technique in Gijón urban area (Northern Spain) at local scale. To accomplish the objective, the observed data of nitrogen oxides, carbon monoxide, sulphur dioxide, ozone and dust for years 2006 to 2008 are used to create a highly nonlinear model of the air quality in the city based on SVM techniques. One objective of this model was to make a preliminary estimate of the dependence between primary and secondary pollutants in Gijón urban area at local scale. A second aim was to determine the factors with the greatest bearing on air quality in order to propose health and lifestyle improvements. This support vector regression model captures the main insight of statistical learning theory to obtain a good prediction of the dependence among the main pollutants in Gijón urban area since it shows a good agreement between the observed and predicted values of pollutants using statistical estimators as correlation coefficients, mean errors and root mean squared errors. Finally, conclusions of this study are exposed.
View