Prediction of Indoor Air Exposure from Outdoor Air Quality Using an Artificial Neural Network Model for Inner City Commercial Buildings

A Review of Artificial Neural Network Models Applied to Predict Indoor Air Quality in Schools

BACKGROUND

Indoor air quality (IAQ) in schools can affect the performance and health of occupants, especially young children. Increased public attention on IAQ during the COVID-19 pandemic and bushfires have boosted the development and application of data-driven models, such as artificial neural networks (ANNs) that can be used to predict levels of pollutants and indoor exposures.

METHODS

This review summarises the types and sources of indoor air pollutants (IAP) and the indicators of IAQ. This is followed by a systematic evaluation of ANNs as predictive models of IAQ in schools, including predictive neural network algorithms and modelling processes. The methods for article selection and inclusion followed a systematic, four-step process: identification, screening, eligibility, and inclusion.

RESULTS

After screening and selection, nine predictive papers were included in this review. Traditional ANNs were used most frequently, while recurrent neural networks (RNNs) models analysed time-series issues such as IAQ better. Meanwhile, current prediction research mainly focused on using indoor PM2.5 and CO2 concentrations as output variables in schools and did not cover common air pollutants. Although studies have highlighted the impact of school building parameters and occupancy parameters on IAQ, it is difficult to incorporate them in predictive models.

CONCLUSIONS

This review presents the current state of IAQ predictive models and identifies the limitations and future research directions for schools.

Estimating Indoor Pollutant Loss Using Mass Balances and Unsupervised Clustering to Recognize Decays

Low-cost air quality monitors are increasingly being deployed in various indoor environments. However, data of high temporal resolution from those sensors are often summarized into a single mean value, with information about pollutant dynamics discarded. Further, low-cost sensors often suffer from limitations such as a lack of absolute accuracy and drift over time. There is a growing interest in utilizing data science and machine learning techniques to overcome those limitations and take full advantage of low-cost sensors. In this study, we developed an unsupervised machine learning model for automatically recognizing decay periods from concentration time series data and estimating pollutant loss rates. The model uses k-means and DBSCAN clustering to extract decays and then mass balance equations to estimate loss rates. Applications on data collected from various environments suggest that the CO₂ loss rate was consistently lower than the PM_2.5 loss rate in the same environment, while both varied spatially and temporally. Further, detailed protocols were established to select optimal model hyperparameters and filter out results with high uncertainty. Overall, this model provides a novel solution to monitoring pollutant removal rates with potentially wide applications such as evaluating filtration and ventilation and characterizing indoor emission sources.

Updating Indoor Air Quality (IAQ) Assessment Screening Levels with Machine Learning Models

Indoor air quality (IAQ) standards have been evolving to improve the overall IAQ situation. To enhance the performances of IAQ screening models using surrogate parameters in identifying unsatisfactory IAQ, and to update the screening models such that they can apply to a new standard, a novel framework for the updating of screening levels, using machine learning methods, is proposed in this study. The classification models employed are Support Vector Machine (SVM) algorithm with different kernel functions (linear, polynomial, radial basis function (RBF) and sigmoid), k-Nearest Neighbors (kNN), Logistic Regression, Decision Tree (DT), Random Forest (RF) and Multilayer Perceptron Artificial Neural Network (MLP-ANN). With carefully selected model hyperparameters, the IAQ assessment made by the models achieved a mean test accuracy of 0.536–0.805 and a maximum test accuracy of 0.807–0.820, indicating that machine learning models are suitable for screening the unsatisfactory IAQ. Further to that, using the updated IAQ standard in Hong Kong as an example, the update of an IAQ screening model against a new IAQ standard was conducted by determining the relative impact ratio of the updated standard to the old standard. Relative impact ratios of 1.1–1.5 were estimated and the corresponding likelihood ratios in the updated scheme were found to be higher than expected due to the tightening of exposure levels in the updated scheme. The presented framework shows the feasibility of updating a machine learning IAQ model when a new standard is being adopted, which shall provide an ultimate method for IAQ assessment prediction that is compatible with all IAQ standards and exposure criteria.

Performing indoor PM2.5 prediction with low-cost data and machine learning

Purpose

The paper aims to develop a particle matter (PM2.5) prediction model for open-plan office space using a variety of data sources. Monitoring of PM2.5 levels is not widely applied in indoor settings. Many reliable methods of monitoring PM2.5 require either time-consuming or expensive equipment, thus making PM2.5 monitoring impractical for many settings. The goal of this paper is to identify possible low-cost, low-effort data sources that building managers can use in combination with machine learning (ML) models to approximate the performance of much more costly monitoring devices.

Design/methodology/approach

This study identified a variety of data sources, including freely available, public data, data from low-cost sensors and data from expensive, high-quality sensors. This study examined a variety of neural network architectures, including traditional artificial neural networks, generalized recurrent neural networks and long short-term memory neural networks as candidates for the prediction model. The authors trained the selected predictive model using this data and identified data sources that can be cheaply combined to approximate more expensive data sources.

Findings

The paper identified combinations of free data sources such as building damper percentages and weather data and low-cost sensors such as Wi-Fi-based occupancy estimator or a Plantower PMS7003 sensor that perform nearly as well as predictions made based on nephelometer data.

Originality/value

This work demonstrates that by combining low-cost sensors and ML, indoor PM2.5 monitoring can be performed at a drastically reduced cost with minimal error compared to more traditional approaches.

A recurrent neural network using historical data to predict time series indoor PM2.5 concentrations for residential buildings

Due to the severe outdoor PM2.5 pollution in China, many people have installed air-cleaning systems in homes. To make the systems run automatically and intelligently, we developed a recurrent neural network (RNN) that uses historical data to predict the future indoor PM2.5 concentration. The RNN architecture includes an autoencoder and a recurrent part. We used data measured in an apartment over the course of an entire year to train and test the RNN. The data include indoor/outdoor PM2.5 concentration, environmental parameters and time of day. By comparing three different input strategies, we found that a strategy employing historical PM2.5 and time of day as inputs performed best. With this strategy, the model can be applied to predict the relatively stable trend of indoor PM2.5 concentration in advance. When the input length is 2 h and the prediction horizon is 30 min, the median prediction error is 8.3 µg/m³ for the whole test set. For times with indoor PM2.5 concentrations between (20,50] µg/m³ and (50,100] µg/m³ , the median prediction error is 8.3 and 9.2 µg/m³ , respectively. The low prediction error between the ground-truth and predicted values shows that the RNN can predict indoor PM2.5 concentrations with satisfactory performance.

Simulation and Analysis of Indoor Air Quality in Florida Using Time Series Regression (TSR) and Artificial Neural Networks (ANN) Models

Exposures to air pollutants have been associated with various acute respiratory diseases and detrimental human health. Analysis and further interpretation of air pollutant patterns are correspondingly important as monitoring them. In the present study, the 24-h and four-month indoor and outdoor PM2.5, PM10, NO2, relative humidity, and temperature were measured simultaneously for a laboratory in Gainesville city, Florida. The indoor PM2.5, PM10, and NO2 concentrations were predicted using multiple linear regression (MLR), time series regression (TSR), and artificial neural networks (ANN) models. The modeling conducted in this study aims to perform a cross comparison study between these models in a symmetric environment. The value of root-mean-square error was improved by 18.33% in comparison with the MLR model. In addition, the value of the coefficient of determination was improved by 24.68%. The ANN model had the best performance and could predict the target air pollutants at 10-min intervals of the studied building with 90% accuracy levels. The TSR model showed slightly better performance compared to the MLR model. These results can be accordingly referred for studies analyzing indoor air quality in similar building types and climate zones.

Predicting PM2.5 in Well-Mixed Indoor Air for a Large Office Building Using Regression and Artificial Neural Network Models

Although the exposure to PM_2.5 has serious health implications, indoor PM_2.5 monitoring is not a widely applied practice. Regulations on the indoor PM_2.5 level and measurement schemes are not well established. Compared to other indoor settings, PM_2.5 prediction models for large office buildings are particularly lacking. In response to these challenges, statistical models were developed in this paper to predict the PM_2.5 concentration in well-mixed indoor air in a commercial office building. The performances of different modeling methods, including multiple linear regression (MLR), partial least squares regression (PLS), distributed lag model (DLM), least absolute shrinkage selector operator (LASSO), simple artificial neural networks (ANN), and long-short term memory (LSTM), were compared. Various combinations of environmental and meteorological parameters were used as predictors. The root-mean-square error (RMSE) of the predicted hourly PM_2.5 was 1.73 μg/m³ for the LSTM model and in the range of 2.20-4.71 μg/m³ for the other models when regulatory ambient PM_2.5 data were used as predictors. The LSTM models outperformed other modeling approaches across the performance metrics used by learning the predictors' temporal patterns. Even without any ambient PM_2.5 information, the developed models still demonstrated relatively high skill in predicting the PM_2.5 levels in well-mixed indoor air.

A Systematic Review of Air Quality Sensors, Guidelines, and Measurement Studies for Indoor Air Quality Management

The existence of indoor air pollutants—such as ozone, carbon monoxide, carbon dioxide, sulfur dioxide, nitrogen dioxide, particulate matter, and total volatile organic compounds—is evidently a critical issue for human health. Over the past decade, various international agencies have continually refined and updated the quantitative air quality guidelines and standards in order to meet the requirements for indoor air quality management. This paper first provides a systematic review of the existing air quality guidelines and standards implemented by different agencies, which include the Ambient Air Quality Standards (NAAQS); the World Health Organization (WHO); the Occupational Safety and Health Administration (OSHA); the American Conference of Governmental Industrial Hygienists (ACGIH); the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE); the National Institute for Occupational Safety and Health (NIOSH); and the California ambient air quality standards (CAAQS). It then adds to this by providing a state-of-art review of the existing low-cost air quality sensor (LCAQS) technologies, and analyzes the corresponding specifications, such as the typical detection range, measurement tolerance or repeatability, data resolution, response time, supply current, and market price. Finally, it briefly reviews a sequence (array) of field measurement studies, which focuses on the technical measurement characteristics and their data analysis approaches.

Assessment of kitchen emissions using a backpropagation neural network model based on urinary hydroxy polycyclic aromatic hydrocarbons

Kitchen emissions are mixed indoor air pollutants with adverse health effects, but the large-scale assessment is limited by costly equipment and survey methods. This study aimed to discuss the application of backpropagation (BP) neural network models in the assessment of kitchen emissions based on the exposure marker. A total of 3686 participants were recruited for the kitchen survey, and their sleep quality was measured by the Pittsburgh sleep quality index (PSQI). After excluding the confounders, 365 participants were selected to assess their urinary hydroxy polycyclic aromatic hydrocarbons (OH-PAHs) concentrations by ultra-high-performance liquid chromatography/tandem mass spectrometry. Two BP neural network models were then set up using the survey and detection data from the 365 participants and used to predict the total urinary OH-PAHs concentrations of all participants. The total urinary OH-PAHs and 1-hydroxy-naphthalene (1-OHNap) concentrations were significantly higher among the 365 participants with poor sleep quality (global PSQI score > 5; P < 0.05). Results from internal and external validation showed that our model has high credibility (model 2). Further, the participants with higher predicted total urinary OH-PAHs concentrations were associated with the global PSQI score of >5 (odds ratio (OR) = 1.284, 95% confidence interval (CI) = 1.082-1.525 for participants with predicted total urinary OH-PAHs concentrations of over 1.897 μg/mmol creatinine in model 1, and OR = 1.467, 95% CI = 1.240-1.735 for participants with predicted total urinary OH-PAHs concentrations of over 2.253 μg/mmol creatinine in model 2) after adjusting for the confounders. Findings suggest that the BP neural network model is suitable for assessing kitchen emissions, and the urinary OH-PAHs concentrations can be taken as the model outlay.

Prediction model for air particulate matter levels in the households of elderly individuals in Hong Kong

Air pollution has shown to cause adverse health effects on mankind. Aging causes functional decline and leaves elderly people more susceptible to health threats associated with air pollution exposure. Elderly spend approximately 80% of their lifetime at home every day. To understand air pollution exposure, indoor air pollutants are the targets for consideration especially for the elderly population. However, indoor air monitoring for epidemiological studies requires a large population, is labor intensive and time consuming. As a result, a prediction model is necessary. For 3 consecutive days in summer and winter, 24-h average of mass concentrations of fine particulate matter (aerodynamic diameter <2.5 μm: PM_2.5) were measured in indoors for 116 households. A PM_2.5 prediction model for elderly households in Hong Kong has been developed by combining ambient PM_2.5 concentrations obtained from land use regression model and questionnaire-elicited information related to the indoor PM_2.5 sources. The fitted linear mixed-effects model is moderately predictive for the observed indoor PM_2.5, with R² = 0.67 (or R² = 0.61 by cross-validation). The model shows indoor PM_2.5 was positively influenced by outdoor PM_2.5 levels. Meteorological factors (e.g. temperature and relative humidity) were related to the indoor PM_2.5 in a relatively complex manner. Congested living areas, opening windows for extended periods for ventilation and use of liquefied petroleum gas for cooking were the factors determining the ultimate indoor air quality. This study aims to provide information about controlling household air quality and can be used for future epidemiological studies associated with indoor air pollution in large population.

Robust Analysis of PM2.5 Concentration Measurements in the Ecuadorian Park La Carolina

In this article, a robust statistical analysis of particulate matter (PM2.5) concentration measurements is carried out. Here, the region chosen for the study was the urban park La Carolina, which is one of the most important in Quito, Ecuador, and is located in the financial center of the city. This park is surrounded by avenues with high traffic, in which shopping centers, businesses, entertainment venues, and homes, among other things, can be found. Therefore, it is important to study air pollution in the region where this urban park is located, in order to contribute to the improvement of the quality of life in the area. The preliminary study presented in this article was focused on the robust estimation of both the central tendency and the dispersion of the PM2.5 concentration measurements carried out in the park and some surrounding streets. To this end, the following estimators were used: (i) for robust location estimation: α-trimmed mean, trimean, and median estimators; and (ii) for robust scale estimation: median absolute deviation, semi interquartile range, biweight midvariance, and estimators based on a subrange. In addition, nonparametric confidence intervals were established, and air pollution levels due to PM2.5 concentrations were classified according to categories established by the Quito Air Quality Index. According to these categories, the results of the analysis showed that neither the streets that border the park nor the park itself are at the Alert level. Finally, it can be said that La Carolina Park is fulfilling its function as an air pollution filter.

Machine learning and statistical models for predicting indoor air quality

Indoor air quality (IAQ), as determined by the concentrations of indoor air pollutants, can be predicted using either physically based mechanistic models or statistical models that are driven by measured data. In comparison with mechanistic models mostly used in unoccupied or scenario-based environments, statistical models have great potential to explore IAQ captured in large measurement campaigns or in real occupied environments. The present study carried out the first literature review of the use of statistical models to predict IAQ. The most commonly used statistical modeling methods were reviewed and their strengths and weaknesses discussed. Thirty-seven publications, in which statistical models were applied to predict IAQ, were identified. These studies were all published in the past decade, indicating the emergence of the awareness and application of machine learning and statistical modeling in the field of IAQ. The concentrations of indoor particulate matter (PM_2.5 and PM₁₀ ) were the most frequently studied parameters, followed by carbon dioxide and radon. The most popular statistical models applied to IAQ were artificial neural networks, multiple linear regression, partial least squares, and decision trees.

Application of an indoor air pollution metamodel to a spatially-distributed housing stock

Estimates of population air pollution exposure typically rely on the outdoor component only, and rarely account for populations spending the majority of their time indoors. Housing is an important modifier of air pollution exposure due to outdoor pollution infiltrating indoors, and the removal of indoor-sourced pollution through active or passive ventilation. Here, we describe the application of an indoor air pollution modelling tool to a spatially distributed housing stock model for England and Wales, developed from Energy Performance Certificate (EPC) data and containing information for approximately 11.5 million dwellings. First, we estimate indoor/outdoor (I/O) ratios and total indoor concentrations of outdoor air pollution for PM_2.5 and NO₂ for all EPC dwellings in London. The potential to estimate concentration from both indoor and outdoor sources is then demonstrated by modelling indoor background CO levels for England and Wales pre- and post-energy efficient adaptation, including heating, cooking, and smoking as internal sources. In London, we predict a median I/O ratio of 0.60 (99% CIs; 0.53-0.73) for outdoor PM_2.5 and 0.41 (99%CIs; 0.34-0.59) for outdoor NO₂; Pearson correlation analysis indicates a greater spatial modification of PM_2.5 exposure by housing (ρ = 0.81) than NO₂ (ρ = 0.88). For the demonstrative CO model, concentrations ranged from 0.4-9.9 ppm (99%CIs)(median = 3.0 ppm) in kitchens and 0.3-25.6 ppm (median = 6.4 ppm) in living rooms. Clusters of elevated indoor concentration are found in urban areas due to higher outdoor concentrations and smaller dwellings with reduced ventilation potential, with an estimated 17.6% increase in the number of living rooms and 63% increase in the number of kitchens exceeding recommended exposure levels following retrofit without additional ventilation. The model has the potential to rapidly calculate indoor pollution exposure across large housing stocks and estimate changes to exposure under different pollution or housing policy scenarios.

Application of an indoor air pollution metamodel to a spatially-distributed housing stock

Estimates of population air pollution exposure typically rely on the outdoor component only, and rarely account for populations spending the majority of their time indoors. Housing is an important modifier of air pollution exposure due to outdoor pollution infiltrating indoors, and the removal of indoor-sourced pollution through active or passive ventilation. Here, we describe the application of an indoor air pollution modelling tool to a spatially distributed housing stock model for England and Wales, developed from Energy Performance Certificate (EPC) data and containing information for approximately 11.5 million dwellings. First, we estimate indoor/outdoor (I/O) ratios and total indoor concentrations of outdoor air pollution for PM_2.5 and NO₂ for all EPC dwellings in London. The potential to estimate concentration from both indoor and outdoor sources is then demonstrated by modelling indoor background CO levels for England and Wales pre- and post-energy efficient adaptation, including heating, cooking, and smoking as internal sources. In London, we predict a median I/O ratio of 0.60 (99% CIs; 0.53-0.73) for outdoor PM_2.5 and 0.41 (99%CIs; 0.34-0.59) for outdoor NO₂; Pearson correlation analysis indicates a greater spatial modification of PM_2.5 exposure by housing (ρ = 0.81) than NO₂ (ρ = 0.88). For the demonstrative CO model, concentrations ranged from 0.4-9.9 ppm (99%CIs)(median = 3.0 ppm) in kitchens and 0.3-25.6 ppm (median = 6.4 ppm) in living rooms. Clusters of elevated indoor concentration are found in urban areas due to higher outdoor concentrations and smaller dwellings with reduced ventilation potential, with an estimated 17.6% increase in the number of living rooms and 63% increase in the number of kitchens exceeding recommended exposure levels following retrofit without additional ventilation. The model has the potential to rapidly calculate indoor pollution exposure across large housing stocks and estimate changes to exposure under different pollution or housing policy scenarios.

Modeling of residential indoor PM2.5 exposure in 37 counties in China

It is critical to estimate the exposure to indoor air pollution of residents spending most of their time in such microenvironments. However, the understanding regarding PM_2.5 exposure in residential indoor environments is very limited. In this study, we collected participants' basic information and time-activity patterns, as well as details of other factors related to indoor air pollution exposure, through questionnaires presented to a large population in 37 counties of China. Continuous monitoring of ambient PM_2.5 concentrations was performed using an environmental fixed-site monitoring network. Residential indoor PM_2.5 concentrations were predicted using a mass balance model based on the data obtained. Evaluation of continuous daily average residential indoor PM_2.5 exposure doses for large populations during winter revealed concentrations ranged from 67 to 195 μg/m³. Finally, differences in residential indoor PM_2.5 exposure between northern and southern China were investigated. The results suggested that residential indoor PM_2.5 concentrations in northern China, associated with heating, were higher than in the south. The established model could be important for improved understanding of human exposure to indoor PM_2.5 air pollution.

A systematic review of data mining and machine learning for air pollution epidemiology

Background

Data measuring airborne pollutants, public health and environmental factors are increasingly being stored and merged. These big datasets offer great potential, but also challenge traditional epidemiological methods. This has motivated the exploration of alternative methods to make predictions, find patterns and extract information. To this end, data mining and machine learning algorithms are increasingly being applied to air pollution epidemiology.

Methods

We conducted a systematic literature review on the application of data mining and machine learning methods in air pollution epidemiology. We carried out our search process in PubMed, the MEDLINE database and Google Scholar. Research articles applying data mining and machine learning methods to air pollution epidemiology were queried and reviewed.

Results

Our search queries resulted in 400 research articles. Our fine-grained analysis employed our inclusion/exclusion criteria to reduce the results to 47 articles, which we separate into three primary areas of interest: 1) source apportionment; 2) forecasting/prediction of air pollution/quality or exposure; and 3) generating hypotheses. Early applications had a preference for artificial neural networks. In more recent work, decision trees, support vector machines, k-means clustering and the APRIORI algorithm have been widely applied. Our survey shows that the majority of the research has been conducted in Europe, China and the USA, and that data mining is becoming an increasingly common tool in environmental health. For potential new directions, we have identified that deep learning and geo-spacial pattern mining are two burgeoning areas of data mining that have good potential for future applications in air pollution epidemiology.

Conclusions

We carried out a systematic review identifying the current trends, challenges and new directions to explore in the application of data mining methods to air pollution epidemiology. This work shows that data mining is increasingly being applied in air pollution epidemiology.

The potential to support air pollution epidemiology continues to grow with advancements in data mining related to temporal and geo-spacial mining, and deep learning. This is further supported by new sensors and storage mediums that enable larger, better quality data. This suggests that many more fruitful applications can be expected in the future.

Indoor PM2.5 exposure affects skin aging manifestation in a Chinese population

Traffic-related air pollution is known to be associated with skin aging manifestations. We previously found that the use of fossil fuels was associated with skin aging, but no direct link between indoor air pollutants and skin aging manifestations has ever been shown. Here we directly measured the indoor PM_2.5 exposure in 30 households in Taizhou, China. Based on the directly measured PM_2.5 exposure and questionnaire data of indoor pollution sources, we built a regression model to predict the PM_2.5 exposure in larger datasets including an initial examination group (N = 874) and a second examination group (N = 1003). We then estimated the association between the PM_2.5 exposure and skin aging manifestations by linear regression. In the initial examination group, we showed that the indoor PM_2.5 exposure levels were positively associated with skin aging manifestation, including score of pigment spots on forehead (12.5% more spots per increase of IQR, P-value 0.0371), and wrinkle on upper lip (7.7% more wrinkle on upper lip per increase of IQR, P-value 0.0218). The results were replicated in the second examination group as well as in the pooled dataset. Our study provided evidence that the indoor PM_2.5 exposure is associated with skin aging manifestation in a Chinese population.