Next Article in Journal
Predicting Cyberbullying Perpetration in US Elementary School Children
Previous Article in Journal
Continuous Glucose Monitoring (CGM) in Sports—A Comparison between a CGM Device and Lab-Based Glucose Analyser under Resting and Exercising Conditions in Athletes
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:

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

Jierui Dong
Nigel Goodman
1,2,3 and
Priyadarsini Rajagopalan
Sustainable Building Innovation Lab., School of Property, Construction and Project Management, RMIT University, Melbourne, VIC 3000, Australia
HEAL National Research Network, Canberra, ACT 2601, Australia
National Centre for Epidemiology and Population Health, The Australian National University, Canberra, ACT 2601, Australia
Author to whom correspondence should be addressed.
Int. J. Environ. Res. Public Health 2023, 20(15), 6441;
Submission received: 19 June 2023 / Revised: 17 July 2023 / Accepted: 22 July 2023 / Published: 25 July 2023


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.

1. Introduction

1.1. Background

Air pollution exacerbated by climate change is a major challenge to public health. For example, heatwaves and wildfires increase human exposure to air pollution and unpredictable health risks [1,2,3,4]. Current research has shown that air pollution is associated with respiratory diseases, such as asthma and allergies [5], inflammatory bowel diseases [6], and lung cancer [7]. The projected impact of wildfires includes a doubling of heat-related deaths, increased hospitalizations for asthma, pneumonia, and cardiovascular effects, and increased mortality and hospitalizations associated with ozone [8]. During wildfires, the combined effect of high temperature and air pollution is more than what would be expected from the sum of their individual effects [9]. Parry et al. [10] demonstrated that air pollution modifies the association between heatwaves and hospital admissions for cardiovascular disease. Meanwhile, the health impacts of climate change are still likely to be underestimated. For example, climate change and global warming modify the availability and distribution of plant-derived allergens such as more intense and prolonged pollen seasons, increasing in the severity and alteration of the seasonality of symptoms of allergic rhinitis and asthma [11].
As a vulnerable population, children are more susceptible to air pollution. Children will likely experience much higher personal exposure, even at the same pollutant concentrations as adults [12], because infants and children inhale and retain larger amounts of air pollution per unit of body weight than adults [13]. Therefore, the National Ambient Air Quality Standards set by the United States Environmental Protection Agency (EPA) set stricter limits on the concentration of carbon monoxide (CO), nitrogen oxides (NOx), particle matter (PM), and sulfur dioxide (SO2) for children [14]. However, the World Health Organization (WHO) found that there are still excess deaths at low concentrations of these pollutants [15,16]. Currie et al. [17] pointed out that lowering the thresholds for acceptable air pollution levels may be a prudent and necessary step toward improving population health, especially among children.
People spend 90% of their time indoors [18], and children spend up to 1075 hours in classrooms each year [19]. Indoor air quality (IAQ) in schools can be further complicated by factors including location, occupancy rates, space allocation, teaching activities, the presence of mould, and ventilation [20,21,22,23,24].Classrooms could be potential important sources of exposure to bioaerosols such as allergens, fungi species, bacteria and viruses especially influenza and SARS-CoV-2. Low levels of exposure might still cause adverse health outcomes [25,26] and spread of epidemic diseases. Another concern is that classrooms are often inadequately ventilated [27,28], which results in health risks and reduces academic performance [29]. Airborne disease transmission of respiratory viruses has drawn significant attention in indoor bioaerosol research since the COVID-19 outbreak. Site-specific strategies are necessary due to the characteristics of school buildings [30], while also taking into account the socio-economic consequences of health and performance effects on children.

1.2. Sources of Indoor Air Pollutants (IAP) in Schools

Similar to other building types (e.g., residential, commercial), there are many different sources of IAP in schools, including CO, black carbon (BC), NOx, SO2, PM, ozone (O3), and volatile organic compounds (VOCs) [31,32]. These can be derived from ambient air pollution, including traffic [33,34,35], factory emissions [36], wood smoke, or from indoor sources, such as furniture [37,38], building materials [39], fragrance products [40,41], and human activities. IAP in schools is also influenced by teaching activities and equipment [42]. Concentrations can depend on the location, local climate, and nearby sources, as well as potential sinks, including the proximity and extent of urban greenspace [43,44]. Compared to other regions of the world, the concentrations of key pollutants (e.g., PM2.5) in Australian cities are generally lower although occasional exceedances in ozone and PM2.5 can occur during extreme pollution events, such as bushfires, dust storms, or heatwaves [45]. In Australia, the 24-hour threshold of PM2.5 defined by the National Environment Protection Measure (Air NEPM) [46] is 25 μg/m3, the same as WHO standards. However, during bushfires, the concentrations of PM2.5 measured in Sydney exceeded 100 μg/m3 and could reach 500 μg/m3, which is four times higher than the guideline [47]. Therefore, IAP caused by extreme climate events needs to be treated with extra attention.

1.3. Influencing Factors

Many factors can affect IAQ, with complex connections between them. These factors can be divided into the physical conditions within a building and its operation. Physical conditions include factors such as materials, room size, window-to-wall ratio, and air tightness of the building envelope, while the operation mainly refers to the ventilation system, such as ventilation method (mechanical and natural ventilation), supply air temperature and humidity, air exchange rate, and human activities [39,48,49]. IAQ in schools is specifically influenced by occupant density and activities. First, as school schedules and students gather, CO2 concentrations subsequently increase. Second, the frequent movement of students in and out of classrooms leads to the re-suspension of airborne particles [20], which makes IAQ more unpredictable. Third, school activities and materials may produce uncommon pollutants, such as chemistry experiments, art tools, chalk dust, and printing [44,50,51].
The effects of these factors are complicated since they interact with each other and have a synergistic effect on IAQ [52]. For example, based on pollutant monitoring data from major EU cities, Spiru et al. [53] found that, compared to naturally ventilated buildings, mechanically ventilated buildings reduce indoor PM2.5 concentrations but increase the concentrations of indoor NO2. Meanwhile, factors such as climate, cultural context, building characteristics and interior design features influence user behaviours [54,55], which also affect IAQ.

1.4. Evaluation of IAQ

Existing research typically uses two types of indicators: indicators based on one or more pollutant concentrations, such as a combination of several different pollutant concentrations, including PM2.5, VOCs and NO2, and indicators of IAP exposure levels or health effects.

1.4.1. Indicators Based on Pollutant Concentrations

One of the most popular indicators is carbon dioxide (CO2) concentration [56,57]. Research on schools showed that CO2 concentration is associated with sick building syndrome (SBS) [58] and affects academic performance [59]. However, there is very limited evidence that CO2 levels below 5000 ppm influence health symptoms, and the results on the effects of moderate CO2 levels on human cognitive performance are not consistent [60]. These findings are likely due to inadequate ventilation elevating the concentrations of other harmful pollutants (e.g., VOCs) along with CO2 levels [61]. Indoor CO2 can be much higher than outdoor mainly due to production by humans, especially in crowded buildings, such as schools. Although the total CO2 generation rate of adults is approximately 18%–54% higher than that of children, the difference between adults and children was not significant at low activity levels, such as sedentary [62]. Meanwhile, many studies used the concentrations of PM2.5, PM10, NO2, O3, CO, VOCs, and fungi to describe IAQ, both in schools and other buildings [21,63,64,65,66,67]. Considering the relationship between pollutant emissions and environmental parameters such as temperature and humidity [68,69], some research added those environmental parameters to IAQ indicators [70,71,72,73,74], while others presented them as contextual factors. However, there is a lack of clear consensus on which or what combination of pollutant parameters should be used to describe IAQ [75].

1.4.2. Indicators of Health Effects and Exposure Levels

Many studies have further focused on the health risks of people exposed to IAP. Carslaw et al. [76] defined a novel method named secondary product creation potential (SPCP) for ranking indoor VOCs in terms of potentially harmful product formation. Although this sequencing can vary with room characteristics and ventilation type, it shows great potential for identifying the potential hazards of VOCs. Baloch et al. [77] used the total VOCs score as the sum of five VOCs (formaldehyde, benzene, naphthalene, limonene and tetrachloroethylene) to elevate the multi-exposure in the classroom. However, this indicator did not cover other common IAPs, such as PM2.5. Taking into account exposure levels, thresholds, and cumulative exposures, López et al. [78] calculated and compared 10 different CO2-based IAQ indicators, including the mean CO2 concentration, the mean concentration above a threshold value, the percentage of time spent in a concentration range, and cumulative exposures greater than a threshold value. However, the exposure limit values (ELVs) for IAP defined by WHO, European Commission, and authorities in 11 countries varied widely, even by order of magnitude for the same averaging exposure periods [75].
For health risks, Logue et al. [79] combined available disease incidence and disease impact models for specific pollutant-disease combinations with data on measured concentrations to estimate the disability-adjusted life-years (DALYs) lost due to inhalation of indoor pollutants. DALYs are a composite indicator of disease and injury burden, capturing both premature mortality and the prevalence and severity of ill health [80]. Furthermore, Cony Renaud Salis et al. [75] proposed three IAQ indicators (Short-Term Exposure Limit, Long-Term Exposure Limit, and IAQ-DALYs) to calculate the acute and chronic effects separately. Another health indicator is peak expiratory flow rates (PEFR), which is commonly used for ambulatory evaluation of asthma [81], as well as air pollutant concentration [82] and exposure levels [83]. For the health risk analysis of IAP exposure, characteristics of the indoor environment and pollutants need to be included, as well as vulnerability, which includes sensitivity (personal characteristics such as age, gender, race, socioeconomic status, pre-existing disease) and adaptive capacity (individual or community-based coping and adapting mechanisms) [84]. Schibuola et al. [85] specifically examined the exposure of children in school by hazard index (HI) and cancer risk (CR), which derived from the US EPA [86]. However, they only calculated the risk from CO2 and PM as the measured concentrations of CO, ozone, and VOCs were below the limits. It probably underestimates the health risk in school, because IAPs such as benzene, polycyclic aromatic hydrocarbons, and ozone, do not have a safe exposure threshold [16,87].

1.5. Machine Learning Predictive Models

Predictive models can help and account for future uncertainties, especially for predicting air pollution patterns and health effects in the context of climate change. Predictive research related to indoor environments involves several scientific fields. Predictive models can be divided into physics-related building simulation (Computational Fluid Dynamics, CFD) and mathematics-related machine learning algorithms [88]. CFD can provide detailed predictions of airflow [89], heat transfer [90] and contaminant transportation and distribution [91]. In addition, CFD can be used to study the spread and propagation of viruses in different scenarios [92]. However, CFD modelling requires detailed information about the indoor space to represent the location digitally using a fine grid, which means it requires substantial computational costs. In contrast, as a data-driven method, machine learning predictive models skip the time-consuming physical equations and derive associations between features from the dataset. Due to fast calculation times and high accuracy, machine learning has been used to assess the built environment, including in models that simulate building loads [93], building energy efficiency [94,95,96], occupancy rates [97], thermal comfort [98], as well as IAQ [99,100]. Wei et al. [99] concluded that the three most commonly used machine learning algorithms for predicting IAQ are Artificial Neural Networks (ANNs), Regression models, and Decision Tree-based models. They compared the effectiveness of those three algorithms simultaneously and found that ANNs were better for predicting nonlinear problems, such as IAQ. This finding has also been confirmed in other research on predicting IAQ [101,102].
ANNs is a data-driven algorithm based on the interconnected structure of neurons. It works on three layers (input layer, hidden layer, and output layer). The predicted output of the neural network is compared with the actual value. Based on the error, the weights assigned to each neuron are changed and then fed into the neural network again until the error is acceptable [103]. Even early ANNs using the back propagation (BP) algorithm with one hidden layer can fit continuous functions of arbitrary complexity [104]. Challoner et al. [105] used ANNs to predict indoor concentrations of PM2.5 and NO2 and found that the prediction of NO2 concentration was more accurate than PM2.5, suggesting that the correlation between input variables and output variables determines the performance of the models. For indicators of non-pollutant concentrations, Xie et al. [106] used ANNs to predict a health-related IAQ indicator, PIAQ, defined as the percentage of occupants in the sampling area of the office building with two or more persistent symptoms calculated for SBS. The input variables included six indoor air pollutants (CO2, PM2.5, HCHO, TVOCs, bacteria, and fungi) and three indoor comfort variables (temperature, relative humidity, and wind speed). With the development of machine learning, an increasing number of ANNs variants, such as convolutional neural networks (CNNs) [107,108] and recurrent neural networks (RNNs) [109,110], are being applied to IAQ research problems. As a data-driven approach, the effectiveness of an ANNs greatly depends on the training dataset. This paper reviews IAQ prediction research using neural network algorithms in schools, including classrooms and daycare centres where children gather. The review summarises data collection and factor selection in modelling processes and explores the status and limitations of IAQ prediction in schools. Finally, this paper discusses how predictive models can enhance IAQ in terms of smart building management and recommends potential research directions for the use of ANNs to study IAQ in schools.

2. Materials and Methods

This review was guided by the protocols outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [111], and included a four-step systematic data collection process (identification, screening, eligibility, and inclusion). As the evaluation was conducted by a single reviewer, we did not register our review, or specifically assess bias among the articles.

2.1. Identification

To carry out this review, the keywords “indoor air quality”, “indoor air pollution”, “predict* OR forecast*”, “neural network”, and “model* OR algorithm*” were used to search in Web of Science, Scopus, and PubMed.

2.2. Initial Screening

In this step, review articles and conference review papers were excluded. Conference proceedings were also excluded as they were on topics unrelated to the intended target of this review.

2.3. Eligibility

In the eligibility phase, research on IAQ of school buildings was expected to be retained. The following criteria were used to exclude studies based on their title and abstract:
  • Studies not directly related to IAQ but examining parameters such as temperature, humidity, occupancy, phytoremediation using plants, outdoor pollution, energy optimization, medical diagnosis, and sensor design;
  • Studies reporting data from residential buildings, commercial buildings, offices, agricultural buildings, industry, underground environments, laboratories, kitchens, and hospitals (i.e., not in schools);
  • Studies focusing on the link between one specific pollution source, such as VOCs (e.g., from furniture emissions), radon (e.g., from rock and soils), NO2 (e.g., from traffic emissions), and indoor levels of pathogens and viruses (e.g., COVID-19);
  • Studies focusing on the effect of IAQ, such as teaching performance and diseases, such as asthma.
Based on the full text, studies where buildings are represented as simple square or two-dimensional models were excluded. These models are difficult to validate and ignore the influence of building factors in IAQ problems. In addition, articles with the same models published by the same authors in multiple journals were screened, and articles focused on IAQ prediction were retained.

3. Results

3.1. Results of the Review Process

Figure 1 depicts the process and results of implementing this review. Publication dates were set before 15 May 2023, and non-English papers were excluded. The search resulted in 134 publications in Web of Science, 160 in Scopus, and 12 in PubMed. After removing duplicate papers, 191 articles remained for screening. In the initial screening, 13 articles were removed, and 178 articles remained for the eligibility phase. Based on the title and abstract, 156 articles were removed in the eligibility step. Then, based on the full text, 13 articles were removed. Finally, nine articles were selected for inclusion in this review.

3.2. Algorithms

Table 1 lists the details of the predictive models in the nine papers. The algorithms used in the articles are summarised, including traditional ANNs and RNNs.

3.2.1. Traditional Artificial Neural Network

Traditional artificial neural networks are designed to emulate neuronal structures in the human brain. It includes an interconnected series of layers, including an input layer, a hidden layer, and an output layer. Each layer is composed of several interconnected nonlinear processing components called neurons or nodes to extract features of the dataset [120]. When input variables are received, the neurons continuously adjust the weights and biases, and iterate according to the error between output and target value until convergence [121]. The dataset is usually divided into a training set, a validation set, and a test set. The training set is for training the weights and biases, while the validation set is for determining when to stop training to avoid overfitting. The test set is used to evaluate the performance of the predictive model. In addition, ANNs predictive models can quantify the extent to which these input variables affect the output variables [122]. For IAQ research, the output is the indicator that describes IAQ (e.g., CO2 and PM2.5 concentration) and the input is the variables that affect IAQ (e.g., outdoor parameters, occupancy rates). A total of eight papers used traditional ANNs to predict IAQ, including research comparing multiple algorithms. These variables are summarised in Section 3.3.3.

3.2.2. Recurrent Neural Network

As one of the variants of ANNs, the structure of RNNs is similar with ANNs, with the difference that the neurons of RNNs also receive their own state information and pass it around. In other words, RNNs neurons receive input variables at this moment and the information of previous data. Therefore, it is particularly suitable for analysing time-series relationships [123], such as NO2 and PM concentrations, which vary due to factors such as human behaviours and seasonal change [124].
However, RNNs with large time spans suffer from the “vanishing and exploding gradient problem” [125]. To address these issues, Hochreiter et al. [126] proposed Long Short-Term Memory (LSTM). The neurons of LSTM are more complex, containing “forget gate”, “input gates”, “output gates”, and cell state functions. The gates function as the controller of the information, i.e., the information to be added or removed is controlled through the gates. The forget gate is used for deciding what information in the cell state to keep or throw away. The input gate helps decide which values are to be updated (i.e., the information we want the network to memorise by cell states). The output gate tells about the information the hidden state will remember [117]. In research comparing neural network algorithms, LSTM achieved the best performance in predicting IAQ [127,128], yet those research were not set in schools. It can be noticed from the model structure that LSTM has more computational parameters compared to RNNs and thus has a slower learning speed [129]. Therefore, Sharma et al. [117] simplified the LSTM structure when predicting CO2 and PM2.5 by removing the forget gate to reduce the execution time.

3.3. Modelling

This section reviews the modelling process of the predictive model, including data collection, factor selection, and variables (Table 2). Due to the different locations, building characteristics, and monitoring conditions, this review focuses on the common features of predictive models for schools, and the differences are also analysed.

3.3.1. Data Collection

Table 1 shows that the data sources include field monitoring and simulation data. For data-driven prediction models, data from actual field sampling provides more information [48]. However, it is usually limited by building conditions and monitoring costs [114], such as monitoring over a large time span or multiple scenarios [118]. Simulation techniques, on the other hand, can obtain data from different scenarios of the same building. For example, Cho et al. [118] set up 241 scenarios of varying HVAC systems, including heating, cooling, heat recovery, recirculation, and ventilation, for a total of 200,000 samples. Moreover, in their subsequent training, the input variables used to predict IAQ included different scenarios and metabolic rates, which are difficult to obtain from field monitoring. However, the simulation models assume and ignore some parameters, such as air change rate and occupancy density, leading to the need for further validation of the effectiveness of such models [118].
The duration of monitoring varies from 7.3 hours [118] to over 1 year [112], which means the data include information on diurnal variations and seasonal changes. Since the routine in school buildings is fixed, the monitoring time or simulation time was chosen for working hours [112,115,117,118].

3.3.2. Factor Selection

To improve the performance of predictive models, factor selection is often used to remove irrelevant factors or redundant factors for training datasets. Factor selection in the nine papers reviewed can be divided into: select indicators influencing IAQ, i.e., output variables, and select input variables for predictive models. In Table 2, Rastogi et al. [112] used Grey Relational Analysis for screening pollutants that have the most significant impact on IAQ. They measured six indoor parameters (temperature, humidity, CO2, CO, PM2.5, PM10) and screened PM2.5 and PM10 as the output of the predictive model. Other papers selected input variables, and the methods included sensitivity tests, correlation analysis, and Principal Component Analysis (PCA). For example, Sharma et al. [117] used the Pearson correlation coefficient to calculate the correlations between pollutants and environmental parameters (temperature, humidity, wind direction, etc.) and then developed models for predicting CO2 and PM2.5, respectively. Elbayoumi et al. [115] used univariate analysis and included ambient parameters, seasonal variation, number of people, and teaching activities as influences on PM concentrations. In this study, sensitivity analysis was used for prediction in different seasons because the weights of the input variables change according to the season. However, teaching activities were not included in input variables due to the limitation of data collection. Hu et al. [116] set four cases with different combinations of input parameters and found the accuracy was better using only historical concentration data. Notably, they also used sensitivity analysis to calculate the time length of the input variables and found a better accuracy using five timesteps (t, t−1…t−4, per hour). In the face of numerous influencing factors, Zhang et al. [48] used PCA to reduce the dimensionality of their dataset, which consisted of 23 factors. They found that over 90% of the variance in the data can be explained by four principal components for both PM2.5, PM10, and NO2, while five principal components were identified for ozone. Moreover, the error of their model with PCA is lower than the traditional predictive model, implying that the original input variables may contain noise.

3.3.3. Variables

Table 2 shows the variables used for training, including output and input. Output variables are the parameters used to represent IAQ. Among the output variables reported in nine reviewed papers, PM2.5 concentration was used in seven sources, CO2 in six sources, and PM10 in four sources. Concentrations of VOCs, NO2, O3, and formaldehyde (HCHO) were also mentioned. Excluding simulation studies, pollutant concentrations were obtained from real-time sensor monitoring. Most models conducted stepwise prediction of pollutants concentrations [113,114,116,117,119] and used multiple variables as outputs, while Rastogi et al. [112] and Elbayoumi et al. [115] used daily average PM2.5 and PM10 as the output variables.
The pollutants and corresponding concentrations reported as outputs in each table are summarised in Table 3. It should be noted that most of the papers, except for number 3 [114] and number 6 [48], used line graphs to represent pollutant concentrations. Therefore, the maximum and minimum values in the table are imprecise. Zhang et al. [48] monitored 11 classrooms, and the maximum and minimum values shown in the table are based on the average concentration in each classroom. During the modelling process, the concentrations of output variables were normalised. However, the model performance outside the extreme range is not validated. For example, the maximum CO2 concentration in [114] is 551 ppm, while in other papers, it exceeded 1000 ppm [112,116,117,119].
As mentioned earlier, numerous influencing factors lead to various input variables. In these nine papers, the input variables can be divided into four categories: indoor parameters, outdoor parameters, building characteristics, occupancy, and time (Table 2). All the papers used indoor parameters as the main input variables, while outdoor data and building characteristics are relatively scarce. For building-related factors, prediction studies face different obstacles in data collection. For example, factors in number 7 (number of fans, room size, and number of floors) are easy to obtain [117], while the natural ventilation rate in number 4 was estimated based on indoor CO2 concentration and number of people [115]. Other interventions that modified the indoor environment were not collected, impacting the model performance. For example, predictive model number 3, excluding the parameters of the HVAC system, resulted in large errors during air conditioning operation [114]. The simulation data compensates for this deficiency. Factors related to mechanical ventilation or air conditioning systems were only used in research number 8, which was trained by simulation data [118]. Due to the availability of data used in the simulation, number 8 was also the only study that used occupancy parameters as input variables. Although the number of people, clothing insulation, and metabolic rate were assumed, this model can be expected to use in real buildings.

3.4. Pollutants Patterns of Predictive Research in Schools

The trends in predicted pollutant levels may be similar within one building; however, it is difficult to make general conclusions for all studies due to the multiplicity of influencing factors, such as location, building structure, and monitoring conditions. Chen et al. [119] predicted CO2, TVOC, and HCHO in different rooms (classrooms, offices, and computer rooms) on the same floor and found TVOC was independent of room function, which meant TVOC data from a single sensor could be applied to other rooms. Patterns of HCHO in different rooms were not correlated; thus, the predictive models need to be trained separately. The patterns of CO2 collected varied widely by room type, i.e., the data of similar rooms can be referenced. CO2 concentrations in offices and computer rooms first increased and then decreased during working hours, while peaks in classroom CO2 occurred several times and at higher concentrations than in other rooms. It confirms the necessity of providing guidelines based on specific patterns of IAP in schools.

4. Discussion and Future

4.1. Limitation of IAQ Prediction in Schools

The vulnerability of children and the health risk of pollution exposure in classrooms are well recognised. However, a limited amount of research has been conducted to predict IAQ for schools, and even less published research has examined specific pollutants in classrooms. For example, further causes of dangerous exposures levels affecting students’ health are those related to the natural emission of radon gas, which typically accumulates in poorly ventilated classrooms, while chemicals substances (i.e., cyanoacrylate, lead, cadmium, nickel) might be contained in school materials [21]. Due to the lack of enforcement guidelines, the predictive models did not focus on specific pollutants in school scenarios. Therefore, the air quality guidelines in schools need to be updated and generalised to develop predictive models.
It should be noted that the impact of extreme weather events was also not discussed as a scenario or variable in the nine papers reviewed. This means that these predictive models were primarily used under conditions of more regular climatic scenarios, and findings are probably not valid for extreme weather events, such as wildfires and dust storms. Two of the nine reviewed papers used outdoor parameters, such as temperature, pollutants concentration, and wind data from a public historical weather database [114,117]. Considering the trend of climate change, the combination of climate or weather prediction models may be more advantageous for IAQ prediction than using historical data. However, there is a research gap in how IAQ prediction models respond to climate hazards and create safe indoor environments.

4.2. Modelling Process

Current neural network models for predicting IAQ vary widely regarding algorithm structure and selection of variables. Most of the research has used traditional ANNs [48,112,113,114,115,117,118,119], while RNNs may be more advantageous for time-series problems [116]. The differences in school location, season, monitoring costs, and the building itself caused inconsistent results in the choice of variables in the nine papers. Most studies used PM2.5 and CO2 concentrations as output variables because existing guidelines and standards already set limits on them. A few studies included NO2 [48], which is mainly derived from daily traffic emissions. On the other hand, the input variables vary more because the influences of indoor pollution sources are more complex and uncertain. For example, indoor pollutant components and concentrations in schools located in high-traffic areas are significantly different from those in suburban areas, resulting in differences in both input and output variables and weights in predictive models. If input variables with low correlation are selected, it will cause data noise and decrease the performance of the model [48].
The modelling process is not yet standardised and is limited by building conditions and data availability. Frequent sampling and long-term monitoring typically mean higher costs, leading to challenges in the widespread use of IAQ predictive models. For data that are difficult to collect, such as personnel data (e.g., clothing insulation, metabolic rate), the corresponding simulation inputs can be used as the training dataset. In addition, the development of an algorithmic framework can help reduce data dependency. For instance, Tariq et al. [130] developed a framework for predicting pollutant concentrations in subway stations by combining transfer learning (TL) and neural networks. The study involved building a pre-training model with a well-measured subway station dataset and subsequently fine-tuning it with lesser data from other subway stations, ultimately providing well-performing prediction models for four subway stations. This TL framework, combined with pre-training and fine-tuning, can reduce the dependence on data collection. However, TL research applied to IAQ prediction is still in the beginning stages.

4.3. Performance and Aim of Predictive Models

The performance of black-box models is usually indicated by the difference between the predicted and actual values. Hu et al. [116] raised the problem of matching computational speed with dynamic control systems; however, their research still focused on the accuracy of the predictive model rather than the environmental benefits of the model. In fact, The models are rarely used in actual buildings due to the constraints of cost and building access, which results in current research being limited by monitoring costs [131] and computational costs [132]. Sharma et al. [117] built an IAQ platform (including monitoring, prediction, and control) by incorporating user ratings to verify the effectiveness of the platform. However, there is no data related to subsequent changes in pollutant concentration. How to calculate the benefits of the prediction model and how to assess the degree of air quality improvement is yet to be studied.
Predictive models can be used for similar rooms in a building to reduce the cost of installing and operating multiple sensors [133]. However, this should be performed with caution as sensors have the capacity to report localised trends that may be underestimated in models, especially for the non-uniform air pollution caused by photocopiers or paints. Furthermore, models can be combined with dynamic control systems to improve IAQ. For example, Hu et al. [116] aimed to avoid unsatisfactory pollutant peaks by combining a predictive model and dynamic control, concluding that the key to solving the system control delay is to ensure that the delay time interval of dynamic prediction is less than the sampling interval of data collection. Warning and control signals require more investment in building management systems. Sharma et al. [117] developed an Android application called IndoAirSense (IAS) for predicting IAQ in classrooms. IAS provides real-time pollutant levels from sensors and prediction of pollutant concentrations, as well as user alerts and suggestions to improve IAQ. However, such an intelligent system that includes monitoring, prediction, and control needs to face not only the problem of prediction accuracy but also the coordination between building functions, such as ventilation system responsiveness, user activities, and outdoor air quality events. In other words, improving IAQ by predictive models requires alignment between predictive model performance and control system.

5. Conclusions

Predictive models are useful in improving IAQ. In this paper, we reviewed studies on the prediction of IAQ in schools based on neural network algorithms. The conclusions of this paper are as follows:
  • IAQ is currently described by the concentration of pollutants, exposure levels and health risks. In predictive models for schools, the concentration of PM2.5, CO2, PM10., VOCs, NO2, O3, and formaldehyde were used to represent IAQ. Other common pollutants, such as CO, SO2, and bioaerosols, were ignored. IAQ predictive models in schools have not yet covered the pollutants mentioned in the guidelines;
  • Traditional ANNs and their variant RNNs were applied to predict IAQ. Traditional ANNs was the most used, while RNNs was more suitable for analysing time-series problems such as predicting IAQ. Algorithmic frameworks, such as TL, have the potential to reduce the data cost of predictive models;
  • Many factors affect IAQ in schools (e.g., ambient environment, schedules, number of occupants, and teaching activities). However, field data were sparse, and occupancy parameters were only used in predictive models trained by simulation data, which requires further validation. Future research could focus on how to use both field data and simulation data to develop predictive models;
  • Climate models and weather forecasts provide data that have a high potential for IAQ prediction, especially for long-term climate change and extreme weather events. Currently, ambient prediction results have not yet been applied to indoor prediction for schools;
  • The concentrations of indoor pollutants may be similar in rooms with comparable functions, pollutant characteristics, and sources, which offers the possibility of reducing the frequency and number of field measurements. Further work is required to assess the correlation between pollutants and the potential for reducing monitoring costs.

Author Contributions

Conceptualization, J.D., N.G. and P.R.; methodology, N.G. and P.R.; writing—original draft preparation, J.D.; writing—review and editing, N.G. and P.R.; supervision, P.R. and N.G. All authors have read and agreed to the published version of the manuscript.


We acknowledge the HEAL (Healthy Environments And Lives) National Research Network, which receives funding from the NHMRC Special Initiative in Human Health and Environmental Change (Grant No. 2008937).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

No new data were created or analyzed in this study. Data sharing is not applicable to this article.


We gratefully acknowledge the Healthy Environments And Lives (HEAL) National Research Network for supporting this study. We also acknowledge RMIT Research Stipend Scholarship for funding the PhD study of J.D.

Conflicts of Interest

The authors declare no conflict of interest.


  1. Zhang, Y.; Workman, A.; Russell, M.A.; Williamson, M.; Pan, H.; Reifels, L. The Long-Term Impact of Bushfires on the Mental Health of Australians: A Systematic Review and Meta-Analysis. Eur. J. Psychotraumatol. 2022, 13, 2087980. [Google Scholar] [CrossRef]
  2. Johnston, F.; Bowman, D. Bushfire Smoke: An Exemplar of Coupled Human and Natural Systems: Bushfire Smoke. Geogr. Res. 2014, 52, 45–54. [Google Scholar] [CrossRef]
  3. Walter, C.M.; Schneider-Futschik, E.K.; Knibbs, L.D.; Irving, L.B. Health Impacts of Bushfire Smoke Exposure in Australia. Respirology 2020, 25, 495–501. [Google Scholar] [CrossRef] [PubMed]
  4. Milton, L.A.; White, A.R. The Potential Impact of Bushfire Smoke on Brain Health. Neurochem. Int. 2020, 139, 104796. [Google Scholar] [CrossRef]
  5. Poole, J.A.; Barnes, C.S.; Demain, J.G.; Bernstein, J.A.; Padukudru, M.A.; Sheehan, W.J.; Fogelbach, G.G.; Wedner, J.; Codina, R.; Levetin, E.; et al. Impact of Weather and Climate Change with Indoor and Outdoor Air Quality in Asthma: A Work Group Report of the AAAAI Environmental Exposure and Respiratory Health Committee. J. Allergy Clin. Immunol. 2019, 143, 1702–1710. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  6. Kaplan, G.G.; Hubbard, J.; Korzenik, J.; Sands, B.E.; Panaccione, R.; Ghosh, S.; Wheeler, A.J.; Villeneuve, P.J. The Inflammatory Bowel Diseases and Ambient Air Pollution: A Novel Association. Am. J. Gastroenterol. 2010, 105, 2412–2419. [Google Scholar] [CrossRef] [Green Version]
  7. Zhang, Y.; Hopke, P.K.; Mandin, C. (Eds.) Handbook of Indoor Air Quality; Springer Nature Singapore: Singapore, 2022; ISBN 9789811676796. [Google Scholar]
  8. Fisk, W.J. Review of Some Effects of Climate Change on Indoor Environmental Quality and Health and Associated No-Regrets Mitigation Measures. Build. Environ. 2015, 86, 70–80. [Google Scholar] [CrossRef] [Green Version]
  9. Shaposhnikov, D.; Revich, B.; Bellander, T.; Bedada, G.B.; Bottai, M.; Kharkova, T.; Kvasha, E.; Lezina, E.; Lind, T.; Semutnikova, E.; et al. Mortality Related to Air Pollution with the Moscow Heat Wave and Wildfire of 2010. Epidemiology 2014, 25, 359–364. [Google Scholar] [CrossRef] [Green Version]
  10. Green, P.; Hayen, Z. Does Particulate Matter Modify the Short-Term Association between Heat Waves and Hospital Admissions for Cardiovascular Diseases in Greater Sydney, Australia? Int. J. Environ. Res. Public Health 2019, 16, 3270. [Google Scholar] [CrossRef] [Green Version]
  11. Eguiluz-Gracia, I.; Mathioudakis, A.G.; Bartel, S.; Vijverberg, S.J.H.; Fuertes, E.; Comberiati, P.; Cai, Y.S.; Tomazic, P.V.; Diamant, Z.; Vestbo, J.; et al. The Need for Clean Air: The Way Air Pollution and Climate Change Affect Allergic Rhinitis and Asthma. Allergy 2020, 75, 2170–2184. [Google Scholar] [CrossRef]
  12. Trasande, L.; Thurston, G.D. The Role of Air Pollution in Asthma and Other Pediatric Morbidities. J. Allergy Clin. Immunol. 2005, 115, 689–699. [Google Scholar] [CrossRef] [PubMed]
  13. University of Queensland, Woolcock Institute of Medical Research. Australian Child Health and Air Pollution Study (ACHAPS); University of Queensland, Woolcock Institute of Medical Research: Brisbane, Australia, 2012. [Google Scholar]
  14. US EPA Organization. NAAQS Table. Available online: (accessed on 17 January 2023).
  15. WHO World Health Organisation-Ambient (Outdoor) Air Pollution. Available online: (accessed on 7 February 2023).
  16. WHO. WHO Global Air Quality Guidelines; WHO: Geneva, Switzerland, 2021; ISBN 978-92-4-003422-8. [Google Scholar]
  17. Currie, J.; Ray, S.H.; Neidell, M. Quasi-Experimental Studies Suggest That Lowering Air Pollution Levels Benefits Infants’ and Children’s Health. Health Aff. 2011, 30, 2391–2399. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  18. US EPA Organization. Indoor Air Quality. Available online: (accessed on 10 July 2023).
  19. Department of Education and Training, Victoria. School Policy Advisory Guide—School Hours. Available online: (accessed on 3 January 2023).
  20. Chithra, V.S.; Shiva Nagendra, S.M. A Review Of Scientific Evidence On Indoor Air Of School Building: Pollutants, Sources, Health Effects And Management. Asian J. Atmos. Environ. 2018, 12, 87–108. [Google Scholar] [CrossRef]
  21. Pulimeno, M.; Piscitelli, P.; Colazzo, S.; Colao, A.; Miani, A. Indoor Air Quality at School and Students’ Performance: Recommendations of the UNESCO Chair on Health Education and Sustainable Development & the Italian Society of Environmental Medicine (SIMA). Health Promot. Perspect. 2020, 10, 169–174. [Google Scholar] [CrossRef] [PubMed]
  22. Bennett, J.; Davy, P.; Trompetter, B.; Wang, Y.; Pierse, N.; Boulic, M.; Phipps, R.; Howden-Chapman, P. Sources of Indoor Air Pollution at a New Zealand Urban Primary School; a Case Study. Atmos. Pollut. Res. 2019, 10, 435–444. [Google Scholar] [CrossRef]
  23. Oliveira, M.; Slezakova, K.; Delerue-Matos, C.; Pereira, M.C.; Morais, S. Children Environmental Exposure to Particulate Matter and Polycyclic Aromatic Hydrocarbons and Biomonitoring in School Environments: A Review on Indoor and Outdoor Exposure Levels, Major Sources and Health Impacts. Environ. Int. 2019, 124, 180–204. [Google Scholar] [CrossRef]
  24. Haddad, S.; Synnefa, A.; Ángel Padilla Marcos, M.; Paolini, R.; Delrue, S.; Prasad, D.; Santamouris, M. On the Potential of Demand-Controlled Ventilation System to Enhance Indoor Air Quality and Thermal Condition in Australian School Classrooms. Energy Build. 2021, 238, 110838. [Google Scholar] [CrossRef]
  25. Zhang, S.; Mumovic, D.; Stamp, S.; Curran, K.; Cooper, E. What Do We Know about Indoor Air Quality of Nurseries? A Review of the Literature. Build. Serv. Eng. Res. Technol. 2021, 42, 603–632. [Google Scholar] [CrossRef]
  26. Madureira, J.; Paciência, I.; Rufo, J.; Ramos, E.; Barros, H.; Teixeira, J.P.; de Oliveira Fernandes, E. Indoor Air Quality in Schools and Its Relationship with Children’s Respiratory Symptoms. Atmos. Environ. 2015, 118, 145–156. [Google Scholar] [CrossRef] [Green Version]
  27. Choo, C.P.; Jalaludin, J. An Overview of Indoor Air Quality and Its Impact on Respiratory Health among Malaysian School-Aged Children. Rev. Environ. Health 2014, 30, 9–18. [Google Scholar] [CrossRef]
  28. Rajagopalan, P.; Andamon, M.M.; Woo, J. Year Long Monitoring of Indoor Air Quality and Ventilation in School Classrooms in Victoria, Australia. Archit. Sci. Rev. 2022, 65, 1–13. [Google Scholar] [CrossRef]
  29. Woo, J.; Rajagopalan, P.; Andamon, M.M. An Evaluation of Measured Indoor Conditions and Student Performance Using D2 Test of Attention. Build. Environ. 2022, 214, 108940. [Google Scholar] [CrossRef]
  30. Sadrizadeh, S.; Yao, R.; Yuan, F.; Awbi, H.; Bahnfleth, W.; Bi, Y.; Cao, G.; Croitoru, C.; de Dear, R.; Haghighat, F.; et al. Indoor Air Quality and Health in Schools: A Critical Review for Developing the Roadmap for the Future School Environment. J. Build. Eng. 2022, 57, 104908. [Google Scholar] [CrossRef]
  31. Aherne, V. Handbook: Indoor Air Quality, 3rd ed.; Australian Building Codes Board: Canberra, Australian, 2021. [Google Scholar]
  32. Milner, J.T.; Dimitroulopoulou, C.; ApSimon, H.M. Indoor Concentrations in Buildings from Sources Outdoors; Annual Report 2004–2005; Atmospheric Dispersion Modelling Liaison Committee (ADMLC): Sheffield, UK, 2005.
  33. Christian, H.; Lester, L.; Trost, S.G.; Schipperijn, J.; Pereira, G.; Franklin, P.; Wheeler, A.J. Traffic Exposure, Air Pollution and Children’s Physical Activity at Early Childhood Education and Care. Int. J. Hyg. Environ. Health 2022, 240, 113885. [Google Scholar] [CrossRef] [PubMed]
  34. Sunyer, J.; Esnaola, M.; Alvarez-Pedrerol, M.; Forns, J.; Rivas, I.; López-Vicente, M.; Suades-González, E.; Foraster, M.; Garcia-Esteban, R.; Basagaña, X.; et al. Association between Traffic-Related Air Pollution in Schools and Cognitive Development in Primary School Children: A Prospective Cohort Study. PLoS Med. 2015, 12, e1001792. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  35. Forns, J.; Dadvand, P.; Esnaola, M.; Alvarez-Pedrerol, M.; López-Vicente, M.; Garcia-Esteban, R.; Cirach, M.; Basagaña, X.; Guxens, M.; Sunyer, J. Longitudinal Association between Air Pollution Exposure at School and Cognitive Development in School Children over a Period of 3.5 Years. Environ. Res. 2017, 159, 416–421. [Google Scholar] [CrossRef]
  36. Liu, D.-R.; Hsu, Y.-K.; Chen, H.-Y.; Jau, H.-J. Air Pollution Prediction Based on Factory-Aware Attentional LSTM Neural Network. Computing 2021, 103, 75–98. [Google Scholar] [CrossRef]
  37. Zhang, R.; Wang, H.; Tan, Y.; Zhang, M.; Zhang, X.; Wang, K.; Ji, W.; Sun, L.; Yu, X.; Zhao, J.; et al. Using a Machine Learning Approach to Predict the Emission Characteristics of VOCs from Furniture. Build. Environ. 2021, 196, 107786. [Google Scholar] [CrossRef]
  38. Zhang, R.; Tan, Y.; Wang, Y.; Wang, H.; Zhang, M.; Liu, J.; Xiong, J. Predicting the Concentrations of VOCs in a Controlled Chamber and an Occupied Classroom via a Deep Learning Approach. Build. Environ. 2022, 207, 108525. [Google Scholar] [CrossRef]
  39. Hadei, M.; Hopke, P.K.; Rafiee, M.; Rastkari, N.; Yarahmadi, M.; Kermani, M.; Shahsavani, A. Indoor and Outdoor Concentrations of BTEX and Formaldehyde in Tehran, Iran: Effects of Building Characteristics and Health Risk Assessment. Environ. Sci. Pollut. Res. 2018, 25, 27423–27437. [Google Scholar] [CrossRef]
  40. Goodman, N.B.; Wheeler, A.J.; Paevere, P.J.; Selleck, P.W.; Cheng, M.; Steinemann, A. Indoor Volatile Organic Compounds at an Australian University. Build. Environ. 2018, 135, 344–351. [Google Scholar] [CrossRef]
  41. Yang, T.; Xiong, J.; Tang, X.; Misztal, P.K. Predicting Indoor Emissions of Cyclic Volatile Methylsiloxanes from the Use of Personal Care Products by University Students. Environ. Sci. Technol. 2018, 52, 14208–14215. [Google Scholar] [CrossRef]
  42. Becerra, J.A.; Lizana, J.; Gil, M.; Barrios-Padura, A.; Blondeau, P.; Chacartegui, R. Identification of Potential Indoor Air Pollutants in Schools. J. Clean. Prod. 2020, 242, 118420. [Google Scholar] [CrossRef]
  43. Mueller, W.; Steinle, S.; Pärkkä, J.; Parmes, E.; Liedes, H.; Kuijpers, E.; Pronk, A.; Sarigiannis, D.; Karakitsios, S.; Chapizanis, D.; et al. Urban Greenspace and the Indoor Environment: Pathways to Health via Indoor Particulate Matter, Noise, and Road Noise Annoyance. Environ. Res. 2020, 180, 108850. [Google Scholar] [CrossRef]
  44. Fonseca Gabriel, M.; Paciência, I.; Felgueiras, F.; Cavaleiro Rufo, J.; Castro Mendes, F.; Farraia, M.; Mourão, Z.; Moreira, A.; de Oliveira Fernandes, E. Environmental Quality in Primary Schools and Related Health Effects in Children. An Overview of Assessments Conducted in the Northern Portugal. Energy Build. 2021, 250, 111305. [Google Scholar] [CrossRef]
  45. Paton-Walsh, C.; Rayner, P.; Simmons, J.; Fiddes, S.L.; Schofield, R.; Bridgman, H.; Beaupark, S.; Broome, R.; Chambers, S.D.; Chang, L.T.-C.; et al. A Clean Air Plan for Sydney: An Overview of the Special Issue on Air Quality in New South Wales. Atmosphere 2019, 10, 774. [Google Scholar] [CrossRef] [Green Version]
  46. Environment National Environment Protection (Ambient Air Quality) Measure. Available online: (accessed on 8 February 2023).
  47. Yu, P.; Xu, R.; Abramson, M.J.; Li, S.; Guo, Y. Bushfires in Australia: A Serious Health Emergency under Climate Change. Lancet Planet. Health 2020, 4, e7–e8. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  48. Zhang, H.; Srinivasan, R.; Yang, X.; Ahrentzen, S.; Coker, E.S.; Alwisy, A. Factors Influencing Indoor Air Pollution in Buildings Using PCA-LMBP Neural Network: A Case Study of a University Campus. Build. Environ. 2022, 225, 109643. [Google Scholar] [CrossRef]
  49. Kempton, L.; Daly, D.; Kokogiannakis, G.; Dewsbury, M. A Rapid Review of the Impact of Increasing Airtightness on Indoor Air Quality. J. Build. Eng. 2022, 57, 104798. [Google Scholar] [CrossRef]
  50. Laiman, R.; He, C.; Mazaheri, M.; Clifford, S.; Salimi, F.; Crilley, L.R.; Megat Mokhtar, M.A.; Morawska, L. Characteristics of Ultrafine Particle Sources and Deposition Rates in Primary School Classrooms. Atmos. Environ. 2014, 94, 28–35. [Google Scholar] [CrossRef] [Green Version]
  51. Mishra, N.; Bartsch, J.; Ayoko, G.A.; Salthammer, T.; Morawska, L. Volatile Organic Compounds: Characteristics, Distribution and Sources in Urban Schools. Atmos. Environ. 2015, 106, 485–491. [Google Scholar] [CrossRef] [Green Version]
  52. Jia, L.-R.; Han, J.; Chen, X.; Li, Q.-Y.; Lee, C.-C.; Fung, Y.-H. Interaction between Thermal Comfort, Indoor Air Quality and Ventilation Energy Consumption of Educational Buildings: A Comprehensive Review. Buildings 2021, 11, 591. [Google Scholar] [CrossRef]
  53. Spiru, P.; Simona, P.L. A Review on Interactions between Energy Performance of the Buildings, Outdoor Air Pollution and the Indoor Air Quality. Energy Procedia 2017, 128, 179–186. [Google Scholar] [CrossRef]
  54. Qiu, Y.; Yang, F.-A.; Lai, W. The Impact of Indoor Air Pollution on Health Outcomes and Cognitive Abilities: Empirical Evidence from China. Popul. Environ. 2019, 40, 388–410. [Google Scholar] [CrossRef]
  55. Yan, D.; O’Brien, W.; Hong, T.; Feng, X.; Burak Gunay, H.; Tahmasebi, F.; Mahdavi, A. Occupant Behavior Modeling for Building Performance Simulation: Current State and Future Challenges. Energy Build. 2015, 107, 264–278. [Google Scholar] [CrossRef] [Green Version]
  56. Li, L.; Zhang, Y.; Fung, J.C.H.; Qu, H.; Lau, A.K.H. A Coupled Computational Fluid Dynamics and Back-Propagation Neural Network-Based Particle Swarm Optimizer Algorithm for Predicting and Optimizing Indoor Air Quality. Build. Environ. 2022, 207, 108533. [Google Scholar] [CrossRef]
  57. Zhu, Y.; Al-Ahmed, S.A.; Shakir, M.Z.; Olszewska, J.I. LSTM-Based IoT-Enabled CO2 Steady-State Forecasting for Indoor Air Quality Monitoring. Electronics 2022, 12, 107. [Google Scholar] [CrossRef]
  58. Dorizas, P.V.; Assimakopoulos, M.-N.; Santamouris, M. A Holistic Approach for the Assessment of the Indoor Environmental Quality, Student Productivity, and Energy Consumption in Primary Schools. Environ. Monit. Assess. 2015, 187, 259. [Google Scholar] [CrossRef]
  59. Wargocki, P.; Porras-Salazar, J.A.; Contreras-Espinoza, S.; Bahnfleth, W. The Relationships between Classroom Air Quality and Children’s Performance in School. Build. Environ. 2020, 173, 106749. [Google Scholar] [CrossRef]
  60. Fisk, W.J.; Wargocki, P.; Zhang, X. Do Indoor CO2 Levels Directly Affect Perceived Air Quality, Health, or Work Performance? ASHRAE J. 2019, 61, 70–77. [Google Scholar]
  61. Persily, A. Indoor Carbon Dioxide Concentrations in Ventilation and Indoor Air Quality Standards. In Proceedings of the Effective Ventilation in High Performance Buildings, Madrid, Spain, 23 September 2015. [Google Scholar]
  62. Wu, Y.; Li, Y.; Gao, S.; Liu, S.; Yin, H.; Zhai, Y. Carbon Dioxide Generation Rates for Children and Adolescents. Build. Environ. 2023, 237, 110310. [Google Scholar] [CrossRef]
  63. Kim, S.; Li, M.; Senick, J.; Mainelis, G. Designing to Engage Children in Monitoring Indoor Air Quality: A Participatory Approach. In Proceedings of the Interaction Design and Children Conference, London, UK, 21–24 June 2020; ACM: New York, NY, USA, 2020; pp. 323–334. [Google Scholar]
  64. Carreiro-Martins, P.; Papoila, A.L.; Caires, I.; Azevedo, S.; Cano, M.M.; Virella, D.; Leiria-Pinto, P.; Teixeira, J.P.; Rosado-Pinto, J.; Annesi-Maesano, I.; et al. Effect of Indoor Air Quality of Day Care Centers in Children with Different Predisposition for Asthma. Pediatr. Allergy Immunol. 2016, 27, 299–306. [Google Scholar] [CrossRef]
  65. Stamatelopoulou, A.; Asimakopoulos, D.N.; Maggos, T. Effects of PM, TVOCs and Comfort Parameters on Indoor Air Quality of Residences with Young Children. Build. Environ. 2019, 150, 233–244. [Google Scholar] [CrossRef]
  66. Schachter, E.N.; Rohr, A.; Habre, R.; Koutrakis, P.; Moshier, E.; Nath, A.; Coull, B.; Grunin, A.; Kattan, M. Indoor Air Pollution and Respiratory Health Effects in Inner City Children with Moderate to Severe Asthma. Air Qual. Atmos. Health 2020, 13, 247–257. [Google Scholar] [CrossRef] [Green Version]
  67. Cabral, J.P.S. Can We Use Indoor Fungi as Bioindicators of Indoor Air Quality? Historical Perspectives and Open Questions. Sci. Total Environ. 2010, 408, 4285–4295. [Google Scholar] [CrossRef] [PubMed]
  68. Mansouri, A.; Wei, W.; Alessandrini, J.-M.; Mandin, C.; Blondeau, P. Impact of Climate Change on Indoor Air Quality: A Review. Int. J. Environ. Res. Public Health 2022, 19, 15616. [Google Scholar] [CrossRef]
  69. Wolkoff, P. Indoor Air Humidity, Air Quality, and Health—An Overview. Int. J. Hyg. Environ. Health 2018, 221, 376–390. [Google Scholar] [CrossRef]
  70. Rastogi, K.; Lohani, D. Context-Aware IoT-Enabled Framework to Analyse and Predict Indoor Air Quality. Intell. Syst. Appl. 2022, 16, 200132. [Google Scholar] [CrossRef]
  71. Ha, Q.P.; Metia, S.; Phung, M.D. Sensing Data Fusion for Enhanced Indoor Air Quality Monitoring. IEEE Sens. J. 2020, 20, 4430–4441. [Google Scholar] [CrossRef] [Green Version]
  72. Lei, L.; Chen, W.; Xue, Y.; Liu, W. A Comprehensive Evaluation Method for Indoor Air Quality of Buildings Based on Rough Sets and a Wavelet Neural Network. Build. Environ. 2019, 162, 106296. [Google Scholar] [CrossRef]
  73. Saad, S.M.; Shakaff, A.Y.M.; Saad, A.R.M.; Yusof, A.M.; Andrew, A.M.; Zakaria, A.; Adom, A.H. Development of Indoor Environmental Index: Air Quality Index and Thermal Comfort Index. In Proceedings of the 11th Asian Conference on Chemical Sensors, Penang, Malaysia, 16–18 November 2015; p. 020043. [Google Scholar]
  74. Ahn, J.; Shin, D.; Kim, K.; Yang, J. Indoor Air Quality Analysis Using Deep Learning with Sensor Data. Sensors 2017, 17, 2476. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  75. Cony Renaud Salis, L.; Abadie, M.; Wargocki, P.; Rode, C. Towards the Definition of Indicators for Assessment of Indoor Air Quality and Energy Performance in Low-Energy Residential Buildings. Energy Build. 2017, 152, 492–502. [Google Scholar] [CrossRef]
  76. Carslaw, N.; Shaw, D. Secondary Product Creation Potential (SPCP): A Metric for Assessing the Potential Impact of Indoor Air Pollution on Human Health. Environ. Sci. Process. Impacts 2019, 21, 1313–1322. [Google Scholar] [CrossRef] [Green Version]
  77. Baloch, R.M.; Maesano, C.N.; Christoffersen, J.; Banerjee, S.; Gabriel, M.; Csobod, É.; de Oliveira Fernandes, E.; Annesi-Maesano, I.; Csobod, É.; Szuppinger, P.; et al. Indoor Air Pollution, Physical and Comfort Parameters Related to Schoolchildren’s Health: Data from the European SINPHONIE Study. Sci. Total Environ. 2020, 739, 139870. [Google Scholar] [CrossRef]
  78. Rueda López, M.J.; Guyot, G.; Golly, B.; Ondarts, M.; Wurtz, F.; Gonze, E. Relevance of CO2-Based IAQ Indicators: Feedback from Long-Term Monitoring of Three Nearly Zero-Energy Houses. J. Build. Eng. 2021, 44, 103350. [Google Scholar] [CrossRef]
  79. Logue, J.M.; Price, P.N.; Sherman, M.H.; Singer, B.C. A Method to Estimate the Chronic Health Impact of Air Pollutants in U.S. Residences. Environ. Health Perspect. 2012, 120, 216–222. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  80. Murray, C.J.L.; Vos, T.; Lozano, R.; Naghavi, M.; Flaxman, A.D.; Michaud, C.; Ezzati, M.; Shibuya, K.; Salomon, J.A.; Abdalla, S.; et al. Disability-Adjusted Life Years (DALYs) for 291 Diseases and Injuries in 21 Regions, 1990–2010: A Systematic Analysis for the Global Burden of Disease Study 2010. Lancet 2012, 380, 2197–2223. [Google Scholar] [CrossRef] [PubMed]
  81. Lahdensuo, A.; Haahtela, T.; Herrala, J.; Kava, T.; Kiviranta, K.; Kuusisto, P.; Peramaki, E.; Poussa, T.; Saarelainen, S.; Svahn, T. Randomised Comparison of Guided Self Management and Traditional Treatment of Asthma over One Year. BMJ 1996, 312, 748–752. [Google Scholar] [CrossRef] [Green Version]
  82. Neas, L.M.; Dockery, D.W.; Koutrakis, P.; Tollerud, D.J.; Speizer, F.E. The Association of Ambient Air Pollution with Twice Daily Peak Expiratory Flow Rate Measurements in Children. Am. J. Epidemiol. 1995, 141, 111–122. [Google Scholar] [CrossRef]
  83. Bae, W.D.; Kim, S.; Park, C.-S.; Alkobaisi, S.; Lee, J.; Seo, W.; Park, J.S.; Park, S.; Lee, S.; Lee, J.W. Performance Improvement of Machine Learning Techniques Predicting the Association of Exacerbation of Peak Expiratory Flow Ratio with Short Term Exposure Level to Indoor Air Quality Using Adult Asthmatics Clustered Data. PLoS ONE 2021, 16, e0244233. [Google Scholar] [CrossRef]
  84. O’Lenick, C.R.; Wilhelmi, O.V.; Michael, R.; Hayden, M.H.; Baniassadi, A.; Wiedinmyer, C.; Monaghan, A.J.; Crank, P.J.; Sailor, D.J. Urban Heat and Air Pollution: A Framework for Integrating Population Vulnerability and Indoor Exposure in Health Risk Analyses. Sci. Total Environ. 2019, 660, 715–723. [Google Scholar] [CrossRef] [PubMed]
  85. Schibuola, L.; Tambani, C. Indoor Environmental Quality Classification of School Environments by Monitoring PM and CO2 Concentration Levels. Atmos. Pollut. Res. 2020, 11, 332–342. [Google Scholar] [CrossRef]
  86. US EPA Organization. 2014 NATA: Technical Support Document. Available online: (accessed on 3 May 2023).
  87. World Health Organization. WHO: Guidelines for Indoor Air Quality: Selected Pollutants; WHO: Copenhagen, Danmark, 2010; ISBN 978-92-890-0213-4. [Google Scholar]
  88. Foucquier, A.; Robert, S.; Suard, F.; Stéphan, L.; Jay, A. State of the Art in Building Modelling and Energy Performances Prediction: A Review. Renew. Sustain. Energy Rev. 2013, 23, 272–288. [Google Scholar] [CrossRef] [Green Version]
  89. Cuce, E.; Sher, F.; Sadiq, H.; Cuce, P.M.; Guclu, T.; Besir, A.B. Sustainable Ventilation Strategies in Buildings: CFD Research. Sustain. Energy Technol. Assess. 2019, 36, 100540. [Google Scholar] [CrossRef]
  90. Song, J.; Meng, X. The Improvement of Ventilation Design in School Buildings Using CFD Simulation. Procedia Eng. 2015, 121, 1475–1481. [Google Scholar] [CrossRef] [Green Version]
  91. Bourdin, D.; Mocho, P.; Desauziers, V.; Plaisance, H. Formaldehyde Emission Behavior of Building Materials: On-Site Measurements and Modeling Approach to Predict Indoor Air Pollution. J. Hazard. Mater. 2014, 280, 164–173. [Google Scholar] [CrossRef] [PubMed]
  92. Mohamadi, F.; Fazeli, A. A Review on Applications of CFD Modeling in COVID-19 Pandemic. Arch. Comput. Methods Eng. 2022, 29, 3567–3586. [Google Scholar] [CrossRef]
  93. Zhang, L.; Wen, J.; Li, Y.; Chen, J.; Ye, Y.; Fu, Y.; Livingood, W. A Review of Machine Learning in Building Load Prediction. Appl. Energy 2021, 285, 116452. [Google Scholar] [CrossRef]
  94. Wang, Z.; Liu, J.; Zhang, Y.; Yuan, H.; Zhang, R.; Srinivasan, R.S. Practical Issues in Implementing Machine-Learning Models for Building Energy Efficiency: Moving beyond Obstacles. Renew. Sustain. Energy Rev. 2021, 143, 110929. [Google Scholar] [CrossRef]
  95. Ardabili, S.F.; Abdilalizadeh, L.; Mako, C.; Torok, B.; Mosavi, A. Systematic Review of Deep Learning and Machine Learning for Building Energy. arXiv 2022, arXiv:2202.12269. [Google Scholar] [CrossRef]
  96. Seyedzadeh, S.; Rahimian, F.P.; Glesk, I.; Roper, M. Machine Learning for Estimation of Building Energy Consumption and Performance: A Review. Vis. Eng. 2018, 6, 5. [Google Scholar] [CrossRef]
  97. Dai, X.; Liu, J.; Zhang, X. A Review of Studies Applying Machine Learning Models to Predict Occupancy and Window-Opening Behaviours in Smart Buildings. Energy Build. 2020, 223, 110159. [Google Scholar] [CrossRef]
  98. Tien, P.W.; Wei, S.; Darkwa, J.; Wood, C.; Calautit, J.K. Machine Learning and Deep Learning Methods for Enhancing Building Energy Efficiency and Indoor Environmental Quality—A Review. Energy AI 2022, 10, 100198. [Google Scholar] [CrossRef]
  99. Wei, W.; Ramalho, O.; Malingre, L.; Sivanantham, S.; Little, J.C.; Mandin, C. Machine Learning and Statistical Models for Predicting Indoor Air Quality. Indoor Air 2019, 29, 704–726. [Google Scholar] [CrossRef] [PubMed]
  100. Wong, L.-T.; Mui, K.-W.; Tsang, T.-W. Updating Indoor Air Quality (IAQ) Assessment Screening Levels with Machine Learning Models. Int. J. Environ. Res. Public Health 2022, 19, 5724. [Google Scholar] [CrossRef]
  101. Zhang, H.; Srinivasan, R.; Yang, X. Simulation and Analysis of Indoor Air Quality in Florida Using Time Series Regression (TSR) and Artificial Neural Networks (ANN) Models. Symmetry 2021, 13, 952. [Google Scholar] [CrossRef]
  102. Taheri, S.; Razban, A. Learning-Based CO2 Concentration Prediction: Application to Indoor Air Quality Control Using Demand-Controlled Ventilation. Build. Environ. 2021, 205, 108164. [Google Scholar] [CrossRef]
  103. Mahesh, B. Machine Learning Algorithms—A Review. Int. J. Sci. Res. 2018, 9, 381–386. [Google Scholar]
  104. Hornik, K.; Stinchcombe, M.; White, H. Multilayer Feedforward Networks Are Universal Approximators. Neural Netw. 1989, 2, 359–366. [Google Scholar] [CrossRef]
  105. Challoner, A.; Pilla, F.; Gill, L. Prediction of Indoor Air Exposure from Outdoor Air Quality Using an Artificial Neural Network Model for Inner City Commercial Buildings. Int. J. Environ. Res. Public Health 2015, 12, 15233–15253. [Google Scholar] [CrossRef]
  106. Xie, H.; Ma, F.; Bai, Q. Prediction of Indoor Air Quality Using Artificial Neural Networks. In Proceedings of the 2009 Fifth International Conference on Natural Computation, Tianjian, China, 14–16 August 2009; pp. 414–418. [Google Scholar]
  107. Wu, A.; Harrou, F.; Dairi, A.; Sun, Y. Machine Learning and Deep Learning-driven Methods for Predicting Ambient Particulate Matters Levels: A Case Study. Concurr. Comput. Pract. Exp. 2022, 34, e7035. [Google Scholar] [CrossRef]
  108. Sonawani, S.; Patil, K. Air Quality Measurement, Prediction and Warning Using Transfer Learning Based IOT System for Ambient Assisted Living. Int. J. Pervasive Comput. Commun. 2023; online ahead of print. [Google Scholar] [CrossRef]
  109. Nurcahyanto, H.; Prihatno, A.T.; Alam, M.M.; Rahman, M.H.; Jahan, I.; Shahjalal, M.D.; Jang, Y.M. Multilevel RNN-Based PM10 Air Quality Prediction for Industrial Internet of Things Applications in Cleanroom Environment. Wirel. Commun. Mob. Comput. 2022, 2022, 1874237. [Google Scholar] [CrossRef]
  110. Dai, X.; Liu, J.; Li, Y. A Recurrent Neural Network Using Historical Data to Predict Time Series Indoor PM2.5 Concentrations for Residential Buildings. Indoor Air 2021, 31, 1228–1237. [Google Scholar] [CrossRef] [PubMed]
  111. Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Akl, E.A.; Brennan, S.E.; et al. The PRISMA 2020 Statement: An Updated Guideline for Reporting Systematic Reviews. Int. J. Surg. 2021, 88, 105906. [Google Scholar] [CrossRef]
  112. Rastogi, K.; Lohani, D.; Acharya, D. An IoT-Based System to Evaluate Indoor Air Pollutants Using Grey Relational Analysis. In Proceedings of the 2020 International Conference on COMmunication Systems & NETworkS (COMSNETS), Bengaluru, India, 7–11 January 2020; pp. 762–767. [Google Scholar]
  113. Kim, J.; Hong, Y.; Seong, N.; Kim, D.D. Assessment of ANN Algorithms for the Concentration Prediction of Indoor Air Pollutants in Child Daycare Centers. Energies 2022, 15, 2654. [Google Scholar] [CrossRef]
  114. Marzouk, M.; Atef, M. Assessment of Indoor Air Quality in Academic Buildings Using IoT and Deep Learning. Sustainability 2022, 14, 7015. [Google Scholar] [CrossRef]
  115. Elbayoumi, M.; Ramli, N.A.; Yusof, N.F.F.M. Development and Comparison of Regression Models and Feedforward Backpropagation Neural Network Models to Predict Seasonal Indoor PM2.5–10 and PM2.5 Concentrations in Naturally Ventilated Schools. Atmos. Pollut. Res. 2015, 6, 1013–1023. [Google Scholar] [CrossRef]
  116. Hu, L.; Fan, N.; Li, J.; Liu, Y. Dynamic Forecasting Model for Indoor Pollutant Concentration Using Recurrent Neural Network. Indoor Built Environ. 2021, 30, 1835–1845. [Google Scholar] [CrossRef]
  117. Sharma, P.K.; Mondal, A.; Jaiswal, S.; Saha, M.; Nandi, S.; De, T.; Saha, S. IndoAirSense: A Framework for Indoor Air Quality Estimation and Forecasting. Atmos. Pollut. Res. 2021, 12, 10–22. [Google Scholar] [CrossRef]
  118. Cho, J.H.; Moon, J.W. Integrated Artificial Neural Network Prediction Model of Indoor Environmental Quality in a School Building. J. Clean. Prod. 2022, 344, 131083. [Google Scholar] [CrossRef]
  119. Chen, S.; Mihara, K.; Wen, J. Time Series Prediction of CO2, TVOC and HCHO Based on Machine Learning at Different Sampling Points. Build. Environ. 2018, 146, 238–246. [Google Scholar] [CrossRef]
  120. Maleki, H.; Sorooshian, A.; Goudarzi, G.; Baboli, Z.; Tahmasebi Birgani, Y.; Rahmati, M. Air Pollution Prediction by Using an Artificial Neural Network Model. Clean Technol. Environ. Policy 2019, 21, 1341–1352. [Google Scholar] [CrossRef] [PubMed]
  121. Agatonovic-Kustrin, S.; Beresford, R. Basic Concepts of Artificial Neural Network (ANN) Modeling and Its Application in Pharmaceutical Research. J. Pharm. Biomed. Anal. 2000, 22, 717–727. [Google Scholar] [CrossRef]
  122. Olden, J.D.; Jackson, D.A. Illuminating the “Black Box”: A Randomization Approach for Understanding Variable Contributions in Artificial Neural Networks. Ecol. Model. 2002, 154, 135–150. [Google Scholar] [CrossRef]
  123. Salehinejad, H.; Sankar, S.; Barfett, J.; Colak, E.; Valaee, S. Recent Advances in Recurrent Neural Networks. arXiv 2018, arXiv:1801.01078. [Google Scholar]
  124. Fuoco, F.; Stabile, L.; Buonanno, G.; Trassiera, C.; Massimo, A.; Russi, A.; Mazaheri, M.; Morawska, L.; Andrade, A. Indoor Air Quality in Naturally Ventilated Italian Classrooms. Atmosphere 2015, 6, 1652–1675. [Google Scholar] [CrossRef] [Green Version]
  125. Lipton, Z.C.; Berkowitz, J.; Elkan, C. A Critical Review of Recurrent Neural Networks for Sequence Learning. arXiv 2015, arXiv:1506.00019. [Google Scholar]
  126. Hochreiter, S.; Schmidhuber, J. Long Short-Term Memory. Neural Comput. 1997, 9, 1735–1780. [Google Scholar] [CrossRef]
  127. Altamirano-Astorga, J.; Santiago-Castillejos, I.-A.; Hernández-Martínez, L.; Roman-Rangel, E. Indoor Air Pollution Forecasting Using Deep Neural Networks. In Proceedings of the Pattern Recognition; Vergara-Villegas, O.O., Cruz-Sánchez, V.G., Sossa-Azuela, J.H., Carrasco-Ochoa, J.A., Martínez-Trinidad, J.F., Olvera-López, J.A., Eds.; Springer International Publishing: Cham, The Netherlands, 2022; pp. 127–136. [Google Scholar]
  128. Lagesse, B.; Wang, S.; Larson, T.V.; Kim, A.A. Predicting PM2.5 in Well-Mixed Indoor Air for a Large Office Building Using Regression and Artificial Neural Network Models. Environ. Sci. Technol. 2020, 54, 15320–15328. [Google Scholar] [CrossRef]
  129. Yu, Y.; Si, X.; Hu, C.; Zhang, J. A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures. Neural Comput. 2019, 31, 1235–1270. [Google Scholar] [CrossRef]
  130. Tariq, S.; Loy-Benitez, J.; Nam, K.; Lee, G.; Kim, M.; Park, D.; Yoo, C. Transfer Learning Driven Sequential Forecasting and Ventilation Control of PM2.5 Associated Health Risk Levels in Underground Public Facilities. J. Hazard. Mater. 2021, 406, 124753. [Google Scholar] [CrossRef] [PubMed]
  131. Lagesse, B.; Wang, S.; Larson, T.V.; Kim, A.A. Performing Indoor PM2.5 Prediction with Low-Cost Data and Machine Learning. Facilities 2022, 40, 495–514. [Google Scholar] [CrossRef]
  132. Thomas, P.; Suhner, M.-C.; Derigent, W. Relearning Procedure to Adapt Pollutant Prediction Neural Model: Choice of Relearning Algorithm. In Proceedings of the 2019 International Joint Conference on Neural Networks (IJCNN), Budapest, Hungary, 14–19 July 2019; pp. 1–8. [Google Scholar]
  133. Kapoor, N.R.; Kumar, A.; Kumar, A.; Kumar, A.; Mohammed, M.A.; Kumar, K.; Kadry, S.; Lim, S. Machine Learning-Based CO2 Prediction for Office Room: A Pilot Study. Wirel. Commun. Mob. Comput. 2022, 2022, 9404807. [Google Scholar] [CrossRef]
Figure 1. Literature review process.
Figure 1. Literature review process.
Ijerph 20 06441 g001
Table 1. Basic information on predictive models in schools.
Table 1. Basic information on predictive models in schools.
#AuthorsYearLocationAlgorithmsData SourceData Duration
1Rastogi et al. [112]2020University, IndiaTraditional ANNsField data14 months,
working hours
2Kim et al. [113]2022Child day care centres, KoreaTraditional ANNsField data1 month
3Marzouk et al. [114]2022University, EgyptTraditional ANNsField data2 months
4Elbayoumi et al. [115]20154 Schools, PalestineTraditional ANNsField data8 months, working hours
5Hu et al. [116]2021School, ChinaRNNsField data27 days
6Zhang et al. [48]2022University, USATraditional ANNsField data2 weeks
7Sharma et al. [117]2021University, IndiaTraditional ANNs, RNNs(LSTM)Field data32 days,
working hours
8Cho et al. [118]2022School, KoreaTraditional ANNsSimulation data7.3 hours
9Chen et al. [119]2018University, SingaporeTraditional ANNsField data15 days
Table 2. Information of modelling process in prediction models.
Table 2. Information of modelling process in prediction models.
# Authors Factor Selection Variables
Indoor Parameters Outdoor Parameters Building Characteristics Occupancy Time
1 *Rastogi et al. [112]Grey Relational AnalysisPM2.5, PM10PM2.5, PM10----
2 *Kim et al. [113]-CO2, PM2.5, VOCsT, H, CO2, PM2.5, VOCs---Hour, day
3Marzouk et al. [114]-CO2, T, HAPT, H, WS, WD---
4Elbayoumi et al. [115]Univariate AnalysisPM2.5, PM2.5-10CO2, RH, CO, PM2.5, PM2.5-10, T, WS, RHVR--
5 *Hu et al. [116]Sensitivity Tests, Correlation CoefficientCO2, PM2.5PM2.5, T, RH, CO2, VOCs ----
6Zhang et al. [48]Principal Component AnalysisPM2.5, PM10, NO2, O3Generate new principal component from the original data
7 *Sharma et al. [117]Pearson Correlation CoefficientCO2, PM2.5NO2, CO, T, HWS, WDNo. of fans, Room size, Floor no.--
8 *Cho et al. [118]Stepwise Linear RegressionPMV, CO2, PM2.5, PM10AP, T, RH, CO2, PM2.5, PM10-Multiple HVAC system parameters PMV, No. of people, Clothing insulation, Metabolic rate-
9 *Chen et al. [119]-CO2, VOCs, HCHOCO2, VOCs, HCHO---Hour, Day, Lag length
T: temperature; H: humidity; RH: relative humidity; VR: ventilation rate; PMV: predicted mean vote; AP: air pressure; WS: wind speed; WD: wind direction; *: use time series data.
Table 3. Pollutants and corresponding concentrations reported as output variables in each of the reviewed papers.
Table 3. Pollutants and corresponding concentrations reported as output variables in each of the reviewed papers.
# Authors Pollutants Concentrations
1 *Rastogi et al. [112]PM2.5<50 μg/m3>300 μg/m3-
PM10<50 μg/m3>350 μg/m3-
2 *Kim et al. [113]CO2500 ppm1500–2000 ppm-
PM2.5<40 μg/m3160 μg/m3-
VOCs-16,000 μg/m3-
3Marzouk et al. [114]CO2393 ppm551 ppm448.7–477.9 ppm
4Elbayoumi et al. [115]PM2.5--104 ± 85 μg/m3
PM2.5–10--350 ± 197 μg/m3
5 *Hu et al. [116]CO2300 ppm1500 ppm-
PM2.5-280 μg/m3-
6Zhang et al. [48]PM2.50.63 ± 0.12 μg/m35.08 ± 3.48 μg/m3-
PM100.65 ± 0.13 μg/m35.45 ± 3.71 μg/m3-
NO27.42 ± 3.26 ppb55.58 ± 2.19 ppb-
O30.03 ± 0.23 ppb26.11 ± 3.5 ppb-
7 *Sharma et al. [117]CO2<200 ppm1200 ppm-
PM2.5<150 μg/m3300 μg/m3-
8 *Cho et al. [118]CO2Simulation data
9 *Chen et al. [119]CO2500 ppm1750 ppm-
VOCs2.5 ppm4 ppm-
*: use time series data.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Dong, J.; Goodman, N.; Rajagopalan, P. A Review of Artificial Neural Network Models Applied to Predict Indoor Air Quality in Schools. Int. J. Environ. Res. Public Health 2023, 20, 6441.

AMA Style

Dong J, Goodman N, Rajagopalan P. A Review of Artificial Neural Network Models Applied to Predict Indoor Air Quality in Schools. International Journal of Environmental Research and Public Health. 2023; 20(15):6441.

Chicago/Turabian Style

Dong, Jierui, Nigel Goodman, and Priyadarsini Rajagopalan. 2023. "A Review of Artificial Neural Network Models Applied to Predict Indoor Air Quality in Schools" International Journal of Environmental Research and Public Health 20, no. 15: 6441.

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop