A Prediction Model for Air Pollution using Artificial Neural Network and Multiple Linear Regression

doi:10.21203/rs.3.rs-3973244/v1

Download PDF

Research Article

A Prediction Model for Air Pollution using Artificial Neural Network and Multiple Linear Regression

https://doi.org/10.21203/rs.3.rs-3973244/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Over the past few decades, air pollution and preventive measures have proven scientifically challenging and the issue is still unending on a worldwide scale. The number of contaminants in the air is increasing daily as a result of the expanding population and the settling of more people in metropolitan regions. They have an impact on people's respiratory and cardiovascular systems, which raises the population's risk of disease and increases mortality. To better enhance public health, several attempts have been made by governmental organizations to comprehend and forecast the Air Quality Index. Without a doubt, the most crucial stage in prediction is the creation of a predictive model of the air quality, which will aid in environmental management and raise public awareness. The most important component of tracking air pollution is air quality prediction. Many methods will be useful in developing an effective model for pollution prediction. The best approach for prediction is to use an artificial neural network. Therefore, this study is conducted by gathering data on air pollutants for the U.P. state cities of Meerut and Ghaziabad and creating an optimum model for the air quality forecast.

Particulate Matter

SO2

NO2

Air Quality Index (AQI)

Artificial Neural Networks (ANN)

Due to the quick exponential population expansion and unfavorable rise in living standards, there is a growing demand for resource usage, which results in a limited supply of resources. Large-scale deforestation is done to fulfill the rising demand for various resources, including land. Anthropogenic activities are linked to the alteration in the atmosphere's chemical composition caused by the uncontrolled fast growth of urbanization and industry. The primary factor contributing to the increase of air pollutants in the environment and the subsequent decline in air quality is the overuse of industrial activity.

One of the primary components that allows life to survive on Earth is air. As a result, harm to the air quality impacts life on Earth. Air pollution is the term for when gaseous, liquid, or solid wastes or by-products contaminate the atmosphere, causing materials to deteriorate, visibility to be reduced, and health risks to humans and the biosphere (Bhavsar 2019). Critical air pollution is found in situations where the weather and geography make it difficult for air to circulate and where a significant portion of the population routinely travels between far-flung parts of the metropolis. Severe measures including factory and school closures and traffic restrictions are needed in response to these disasters. In fact, a number of epidemiological investigations have repeatedly demonstrated a link between respiratory and cardiovascular disorders and air pollution. Predicting such occurrences up to two days ahead of time would enable the implementation of more effective preventative measures to protect the public's health (Boznar et al. 1993).

The main focus of this study is the lack of effective prediction tools for making the most use of existing data on the various contaminants found in the air, despite the fact that these data are accessible from multiple pollution control boards. Effective regulations may be put in place to stop this expansion if the change in these factors is forecast with efficiency. This may be achieved by creating an effective algorithm that makes use of the Artificial Neural Network's predictive power to produce highly dependable outcomes.

The location and details pertaining to the sample data collection are given in section II that follows. The procedures for the suggested model and the data preparation is covered in Section III, the findings are covered in Section IV, which concludes.

The data for this project is taken from Pollution Control board, Uttar Pradesh and the sites are assigned in Meerut and Ghaziabad for the forecast of air quality. Because of urbanization and rapid deforestation, it has been seen in recent years that the temperature is rising and the amount of rainfall is falling.

In Uttar Pradesh, the most industrialized districts are Meerut and Ghaziabad. There are several other engineering businesses in addition to the various companies that make consumer products, electrical goods, and fabrication. The decline in air quality is caused by a large concentration of industry, congested transportation, and human habitation. In order to anticipate air quality, the three main air pollutants PM10, SO2 and NO2 are chosen.

The pollution control board's data has to be normalized because it varies greatly over time. The data is normalized within the interval [0,1]. The purpose of the normalization procedure is to make the data less error-prone for training and testing. Normalization may be achieved mathematically by using the formula,

Y= (X-Xmin)/ (Xmax-Xmin).

The variable's value is denoted by X in this formula, whereas Y represents the normalized value. The variable's highest value is represented by Xmax, while its minimum value is represented by Xmin.

An artificial neural network and multiple linear regression methods are used in the development of the air quality forecasting model. Given that the parameter data is a time series, a neural network and multiple linear regression (MLR) for time series data development are necessary. In this study, the input parameters are PM10, SO2, NO2 and the output parameter is AQI. An independent data set is needed for the model's training, testing, and validation in order to construct the ANN and MLR models for air quality prediction.

Consequently, five years of data 2019–2023 is taken for Kesarganj road Meerut and Khora Colony Ghaziabad from the website of the Uttar Pradesh Pollution Control Board.

3.1 Statistical Method

One of the most used statistical methods is regression analysis. A multivariate statistical method called multiple regression analysis is used to look at the connection between one dependent variable and a group of independent variables. The purpose of multiple regression analysis is to use independent variables whose values are known to predict the single dependent variable. The regression or response function f represents the influence of independent factors on the response mathematically:

$y$ - dependent variable.

${b}_{1},{b}_{2},{b}_{3},\dots .,{b}_{n}$ - regression parameters(unknown)

$f-$ It is often believed that the form is known

For the observed response variable, a regression model is developed

$$\text{z}=\text{y}+\text{Ɛ}={f(x}_{1},{x}_{2},{x}_{3,\dots ..,}{x}_{n};{b}_{1},{b}_{2},{b}_{3},\dots .,{b}_{n})+\text{Ɛ}$$

$$\text{Ɛ}- \text{e}\text{r}\text{r}\text{o}\text{r} \text{i}\text{n} \text{z}.$$

The method of least squares can be used to determine unknown regression parameters${b}_{1},{b}_{2},{b}_{3},\dots .,{b}_{n}$

$${E(b}_{1},{b}_{2},{b}_{3},\dots .,{b}_{n})=\sum _{j=1}^{n}{\left({z}_{j}-{y}_{j}\right)}^{2}=\sum _{j=1}^{n}{\left({z}_{j}-{f(x}_{1},{x}_{2},{x}_{3,\dots ..,}{x}_{n};{b}_{1},{b}_{2},{b}_{3},\dots .,{b}_{n})\right)}^{2}$$

Where ${E(b}_{1},{b}_{2},{b}_{3},\dots .,{b}_{n})$ is the error function.

In order to estimate ${b}_{1,}{b}_{2},{b}_{3},\dots .,{b}_{n}$, we solve the system of equations to minimize E:

$$\frac{\partial E}{\partial {b}_{i}}=0, i=\text{1,2},,\dots ,n$$

Values of different parameters in MLR model for Meerut city and Ghaziabad city are given in Table 1 and Table 2 respectively. The graphs of frequency and regression standardized residual are given in Fig. 1 and Fig. 2 for Meerut and Ghaziabad city respectively.

Table 1

Model for Meerut City
Model	R	R Square	Adjusted R Square	Standard Error of the Estimate	R Square Change	F Change	df1	df2	Sig. F Change
1	0.998	0.996	0.996	2.64435	0.996	3861.144	3	47	$<0.001$
Predictors: (Constant), NO2, SO2, PM10 Dependent Variable: AQI

Table 2

Model for Ghaziabad City
Model	R	R Square	Adjusted R Square	Standard Error of the Estimate	R Square Change	F Change	df1	df2	Sig. F Change
1	.995	.989	.988	5.95011	.989	1271.044	3	42	< 0.001

3.2 Multilayer Perceptron

An artificial neural network is a computational or mathematical model that draws inspiration from the architecture and functionality of biological brain networks (Kumar and Sharma 2017). In general, their neuronal activation features, training approach, network structure, connection pattern, and data processing capability define them. The Multilayer Perceptron is the most widely used neural network model. Because this kind of neural network needs a desired output in order to learn, it is referred to as a supervised network. Using historical data, the objective of this kind of network is to build a model that accurately connects the input and output. In this manner, if the desired outcome is unknown, the model can be utilized to produce the output. When the inputs are transferred from the input layer to the hidden layer, they are multiplied by the interconnection weights. They are added together and then handled by a nonlinear function (often the hyperbolic tangent) inside the hidden layer. If there are many hidden layers, the processed data is increased by the connection weights once more when it exits the first hidden layer, summed, and then handled by the second hidden layer, and so forth.

The output of the neural network is finally produced by multiplying the data by the connectivity weights and processing it one more time in the output layer. To train the neural network for any task, a number of input-output mapping experiments are needed. For any trained neural network to produce dependable results, these data are among the most crucial components. Consequently, in order to include all the necessary information, the training sample data must be fairly large and comprise a wide range of data from various process parameters and experimental settings. Figures 3 and 4 shows the structure of neural network for Meerut and Ghaziabad city respectively with inputs PM10, SO2 and NO2, one hidden layer and output as AQI.

Figures 4 and 5 show the modal summary with relative error in training and testing for both cities.

From the above two methods we have shown that the MLR method gives 99.6 percent correctness while multilayer perceptron (MLP) gives 99.9 percent correctness in training and 99.7 percent correctness in testing for Meerut City. Also for the taken data of Ghaziabad city, we have applied both the methods MLR and ANN. In the MLR method, the accuracy is 98.9% but in ANN accuracy is 99.5% in training and 99.4% in testing. So it can be said that Artificial Neural Networks are better for this data in comparison to the MLR method. Neural network science is a very broad discipline with much room for further research in areas such as application, architectural selection, and data pre-processing and representation. The research's next natural step is to enhance neural networks' performance even further for this use case. This might be done by using better input, better design choices, or better training techniques.

* Ethics approval and consent to participate -YES, I APPROVE AND GIVE CONSENT

* Consent for publication -YES, I GIVE CONSENT

* Availability of data and material - http://www.uppcb.com/ambient_quality.htm

* Competing interests -NOT APPLICABLE

* Funding -NOT APPLICABLE

* Authors Contributions -Mr. Lokesh Kumar writes the paper and Dr. Gaurav Kumar do corrections.

Ahmad AL, Azid IA, Yusof AR, Seetharamu KN (2004) Emission control in palm oil mills using artificial neural network and genetic algorithm, Computers and Chemical Engineering, 28, pp. 2709–2715
Asadollahfardi G, Zangooei H, Aria SH (2016) Predicting PM2.5 Concentrations using Artificial Neural Networks and Markov Chain, a Case Study Keraj City. Asian J Atmospheric Environ 10(2):67–79
Bhavsar R (2019) Air Pollution Monitoring Using Artificial Neural Network. Int J Sci Eng Res 10(12):515–519
Boznar M, Lesjak M, Mlakar P (1993) A neural network-based method for short-term predictions of ambient So2 concentrations in highly polluted industrial areas of complex terrain. Atmos Environ 27B:221–230
Boznar MZ, Mlakar P (2002) Use of neural networks in the field of air pollution modeling. Air Pollution Model Its Application XV, pp. 375–383
Chelani AB, Raoi CV, Phadke KM, Hasan MZ (2002) Prediction of sulfur dioxide concentration using artificial neural networks. Environ Model Softw 17:161–168
Cogliani E (2001) Air pollution forecast in cities by an air pollution index highly correlated with meteorological variables. Atmos Environ 35:2871–2877
Comrie AC (1997) Comparing Neural Networks and Regression Models for Ozone Forecasting. Air Waste Manage Association 47:653–663
Elminir HK, Galil HA (2006) Estimation of air pollutant concentrations from meteorological parameters using artificial neural network. J Electr Eng 57:105–110
Freeman AM, III Air pollution and property values, a further comment. Rev Econ Stat, 56, pp. 554– 556, Nov. 1974
Gardner MW, Dorling SR (1999) Neural network modelling and prediction of hourly NOx and NO2 concentrations in urban air in London. Atmos Environ 33:709–719
Guo C, Liu G, Chen CH (2020) Air Pollution Concentration Forecast Method Based on the Deep Ensemble Neural Network, Hindawi Wireless Communications and Mobile Computing, Vol. 2020, Article ID 8854649, 13 pages
Hadjiiski L, Hopke P (2000) Application of artificial neural networks to modeling and prediction of ambient ozone concentrations. J Air Waste Manage Association 50:894–901
Hall T, Brooks HE, Doswell CA (1999) Precipitation forecasting using a neural network. Weather Forecast 14:338–345
liadis LS, Spartalis SI, Paschalidou AK, Kassomenos P (2007) Artificial Neural Network Modelling of the surface Ozone Concentration. Int J Comput Appl Math 2:125–138
Jef H, Clemens M, Gerwin D, Frans F, Olivier B (2005) A neural network forecast for daily average PM10 concentrations in Belgium. Atmos Environ 39:3279–3289
Kumar G, Sharma RK (2017) Air Pollution Evaluation Methods. Int J Eng Res Dev 13(9):12–17
Kumar G (2018) Time series analysis of PM10 for Bulandhshahr Industrial Area in NCR using Multiple Linear Regression. Int J Eng Res Dev 14(3):56–62
Kumar G (2018) Time series analysis of PM10 for Noida Sector 1 Industrial Area in NCR using Multiple Linear Regression. Bull Pure Appl Sci Sect E-Math Stat 37(2):273–277
Lu WZ, Wang WJ, Fan HY, Leung AYT, Xu ZB, Lo SM, Wong JCK (2002) Prediction of Pollutant Levels in Causeway Bay Area of Hong Kong Using an Improved Neural Network Model. J Environ Eng 128:1146–1157
Nagendra SMS, Khare M (2005) Modelling urban air quality using artificial neural network. Clean Technol Environ Policy 7:116–126
Niharika VM, Rao PS (2014) A survey on Air Quality forecasting Techniques. Int J Comput Sci Inform Technol 5(1):103–107
Niska H, Hiltunen T, Karppinen A, Ruuskanen J, Mikko Kolehmainen M (2004) Evolving the neural network model for forecasting air pollution time series. Eng Appl Artif Intell 17:159–167
Nunnari G, Dorling S, Schlink U, Cawley G, Foxall R, Chatterton T (2004) Modelling SO2 concentration at a point with statistical approaches. Environ Model Softw 19:887–905
Prachi KN, Matta G (2011) Artificial neural network applications in air quality monitoring and management. Int J Environ Rehabilitation Conserv 2(1):30–64
Reshma J (2020) Analysis and Prediction of Air Quality. Int Res J Eng Technol 7(1):266–270
Saxena A, Verma N, Tripathi KC (2013) A Review Study of Weather Forecasting Using Artificial Neural Network Approach. Int J Eng Res Technol 2(11):2029–2035
Selvaraj RS, Elampari K, Gayathri R, Jeyakumar SJ (2010) A neural network model for short term prediction of surface ozone at tropical city. Int J Eng Sci Technol 2:5306–5312
Sharma V, Rai S, Dev A (2012) A Comprehensive Study of Artificial Neural Networks. Int J Adv Res Comput Sci Softw Eng 2(10):278–284
Shi JP, Harrison RM (1997) Regression modelling of hourly NOx and NO2 concentrations in urban air in London. Atmos Environ 31:4081–4094
Werner H, Obach M (2001) New neural network types estimating the accuracy of response for ecological modeling. Ecol Model 146:289–298
Zaefizadeh M, Khayatnezhad M, Gholamin R (2011) Comparison of Multiple Linear Regressions (MLR) and Artificial Neural Network (ANN) in Predicting the Yield Using its Components in the Hulless Barley. American-Eurasian J Agricultural Environ Sci 10(1):60–64
SPSS27 software, SPSS Inc., http://www.spss.com

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

A Prediction Model for Air Pollution using Artificial Neural Network and Multiple Linear Regression

Status:

Version 1

Abstract

Figures

I. INTRODUCTION

II. IDENTIFICATION AND COLLECTION OF THE DATA FROM THE MONITORING STATION

III. STEPS INVOLVED IN THE DEVELOPMENT OF MODEL

IV. CONCLUSION

Declarations

References

Additional Declarations

Status:

Version 1