International Journal of Scientific Research and Engineering Development-– Volume 6 Issue 1, Jan-Feb 2023

Available at www.ijsred.com
SEARCH ARTICLE
RESEARCH ARTICLE OPEN ACCESS
OPEN ACCESS

Monitoring and Prediction of Air Pollution Using Machine

Learning Models: A Review
Hemangi Tamore, Brinda Temkar, Sanya Wakode, Sunidhi Yadav, Dr.Jyoti Dange
EXTC, Atharva College Of Engineering, Mumbai
Email: [email protected]
EXTC, Atharva College Of Engineering, Mumbai
Email: [email protected]
EXTC, Atharva College Of Engineering, Mumbai
Email: [email protected]
EXTC, Atharva College Of Engineering, Mumbai
Email: [email protected]
EXTC, Atharva College Of Engineering, Mumbai
Email: [email protected]
-----------------------------------------************************---------------------------------
Abstract:
The influence of Machine learning and Data Science is advancing in healthcare, personalized
recommendation models, environmental studies as well as in education institutes. It has become an
important factor of consideration for both companies as well as individuals. Prediction of air quality is one
of the fields that machine learning has given its contributions. Air quality index measures the
concentration of various gases like carbon dioxide, carbon monoxide, nitrogen dioxide, sulfur dioxide,
particulate matter like smoke, soot, methane that releases after burning natural gas, coal, wood, etc. High
concentrations of these substances can cause severe diseases like lung cancer and even premature deaths.
Machine learning helps in predicting the air quality so necessary actions can be taken if the pollutants
increase more than a certain limit. If pollution of air is not handled carefully and sensibly, some day it can
lead to extinction of humans. This paper provides a revision on the results obtained by researchers on
monitoring and prediction of air pollution using machine learning and IOT. After thorough reviewing it
was observed that the machine learning algorithms used for the analysis were quite effective.

Keywords — Air pollution, air quality prediction, machine learning, iot, regression, maps
-----------------------------------------************************---------------------------------
INTRODUCTION sulfur dioxide(SO2), Ozone (O3), nitrogen oxide
Human society is evolving at a great pace. This is (NO) and nitrogen dioxide (NO2) and a complex
evident in every city, including various problems mixture of solid and liquid droplets called
caused due to these evolutions and expansions. The Particulate Matters (PM, e.g., PM2.5, PM10) [2].
rapid growth in cities naturally increases pollution There are many studies which show that respiratory
and pollutants caused due to heavy traffic, disorders are due to air pollution. A study
industries, etc. Because of this, the quality of air is conducted by State of Global Air (SOGA) shows
polluted and has a great impact on human health. that continuous exposure to air pollution can reduce
There are certain pollutants which affect human life expectancy by up to 20 months [2]. Air
health and have a great impact these pollutants are, pollution is accountable for the death of 7 million
carbon dioxide (CO2), carbon monoxide (CO), persons worldwide each year or one in eight

ISSN : 2581-7175 ©IJSRED: All Rights are Reserved Page 156

International Journal of Scientific Researc
earch and Engineering Development-– Volume 6 Issue
Issu 1, Jan-Feb 2023
Available at www.ijsred.com
premature deaths yearly [4]. Hence, it is a difficult I. LITERATURE REVIEW
task to predict the quality of air in a particular In [1] the research proposedd a Web application
region, which makes people less caref reful about the based on air pollution monito itoring systems using
existing air quality. The World Healthh Organisation HAPS-based wireless sensorr networks that were
(WHO) set up guidelines and limitatio tions for urban able to display information abobout air grade data in
air pollution that should be respected ed in order to graphical and table form. Thi his is in accordance
protect the citizens from the pollutants ts [2]. Almost with the design of the web ebsite where the air
570,000 children under the age of five fiv die every condition menu contains a lineine-shaped graph that
year from respiratory infection linked to is divided into two sub menu nus: graphs CO and
indoor/outdoor pollution and second--hand smoke PM10 levels.
[4]. Small children when made to com ome in contact
with air pollution may have a risk of developing [2] This paper presents an IoToT Platform covering
respiratory disorders. Therefore, a needed to develop a the data collection process from
rom the sensing nodes
system that predicts air quality aaccurately is to visualization for the end pro
rocess. They provide
necessary. an easy to assemble and du duplicate design for
Air Quality Index(AQI) is calculated on the basis of research purposes to ena nable fast sensor
concentration of different pollutants likike CO2, SO2, deployment and data collectiotion. It is a low level
CO, NH3, PM2.5, PM10, etc. as shownn in Fig. 1 the middle- ware, meaning that it ccan be implemented
index uses different colors for better understanding
u directly on top of the phys ysical layer of low
of the level of pollutants in air.Here, green
gre indicates consumption protocols such ch as the Zigbee
least polluted air while maroon indica cates the most protocol.
polluted air.
In [3] the paper proposes thatt using
u a multivariate
modeling approach enhance ces the prediction
accuracy and reduces error ror because of the
dependency between targett gasses and other
features included such as tempmperature, day of the
week, and H2S.

In [4]this paper the system waas designed using an

Fig. 1. Air Quality Index(AQI-iindex) Arduino microcontroller.Thhis system was
This paper includes Section 2 which describes
d the develped to monitor and analyzlyze real time data air
methodology adopted by differentt researchers, quality and log data to a rememote server,the data
Section 3 explores the results obtained
ed on the same kept updated over the in internet.Air quality
and finally Section 4 includes the concl
clusion. measurement were taken usin sing Part per Million
matrics(PPM) and analyzee using Microsoft
Excel.This designed system w was taken accurate
measurement of Air quality.T y.The accurate result
was displayed on the designed ed hardwares display
interface and it could be access
ssed via cloud on any
smart phone service.

[5]This paper has used Machine learning

techniques to predict concentr
ntration of the so2 in
Fig. 2. General steps in predicting Air
A Quality
the environment. Sulfur dioxid
ide effects on the skin
Prediction and mucous membraranes of the

©IJS
©IJSRED: All Rights are Reserved Page
International Journal of Scientific Research and Engineering Development-– Volume 6 Issue 1, Jan-Feb 2023
Available at www.ijsred.com
eyes,nose,throat,and lung. This system employed [10]This paper has determined PM10 level best
models in Time series to predict so2 in the with Random Forest but does not accurately predict
environment.In this paper they have used Time the level of dangerous pollutants but can work with
series forecast models for prediction of Air incomplete data sets.
Quality Index in Metro cities.
III.COMPARATIVE STUDY AND ANALYSIS
In [6]this SARIMAX and Holt-Winter’s models
are used to predict the air quality index. These time
In this section we observe the performance of
series forecasting models can be utilized to predict
the values of the Air Quality Index(AQI) based on different machine learning models, adopted by
past data. To analyze the performance of models researchers in their studies which are as follows:
Mean Absolute Percentage Error (MAPE) is used as
the score function. The Holt-Winters algorithm can Table I: Performance of Different ML Models
handle seasonality,but results produced by this
ML SO2 CO O3 NO PM PM
model have not very accuracy. On the other
Model 2 2.5 10
hand,The SARIMAX can handle seasonality and
has much better result accuracy than the Holt- Linear 0.12 0.02 0.09 0.1 0.02 0.02
Winters model. Regressi 5
on
In [7]this paper,they have compared the decision
tree, linear regression and random forest. Major air Decision 0.80 0.61 0.62 0.6 0.75 0.61
pollutants are taken and meteorological conditions Tree 60 4
are taken using the Arduino Platform. Random
forest gives better accurate results due to overfitting Random 0.85 0.79 0.79 0.7 0.86 0.79
that reduces errors But drawback is Random forest Forest 6 01
uses more memory and high cost. Haotian Jing & regressio
Yingchun Wang(2020). n

[8]had predicted the air quality index using XG SARIM 0.86 0.75 0.83 0.8 0.72 0.81
Boost.It uses the weak classifier and shortcoming of AX 3 2 66
the previous weak classifier to form a strong
classifier thus reducing the error between predicted
and actual values. It uses the K- cross The above comparison was obtained by observing
validation.The mean absolute error and coefficient the results and performance parameters of the
of determination is determined to predict the mentioned algorithms in [6] and [7].
difference between actual and predicted value. The
drawback is that it takes the previous value and is
affected by outlier unwanted pollutants in the air. IV. CONCLUSION
In this paper, we have successfully understood the
[9]Their research had predicted the level of air need of machine learning and the huge role that it
pollutant with Recurrent neural network at any
plays in prediction of air pollution. The papers
given time and removed the drawback of hourly
prediction due to memorization power of algorithm discussed above proposed Air Pollution
but lacks in working without memory operation. Monitoring and Prediction systems using Machine
Learning Models out of which a few successfully
predicted the pollution levels accurately while

ISSN : 2581-7175 ©IJSRED: All Rights are Reserved Page 158

International Journal of Scientific Research and Engineering Development-– Volume 6 Issue 1, Jan-Feb 2023
Available at www.ijsred.com
others didn’t. After going through these papers we [12] C. Santos, J. A. Jiménez, and F. Espinosa, “Effect of Event-
Based Sensing on IOT Node Power Efficiency. Case Study: Air
observed that SARIMAX was the best proposed Quality Monitoring in Smart Cities,” IEEE Access, vol. 7, pp.
model as it took the parameter of seasonality into 132577–132586, 2019.
[13] D. Wei, “Predicting air pollution level in a specific city,” 2014.
consideration which helped in correctly predicting [14] Kostandina Veljanovska and Angel Dimoski, “Air Quality
the air quality that would then alert potential users Index Prediction Using Simple Machine Learning Algorithms,”
International Journal of Emerging Trends & Technology in
and also help in future predictions. Computer Science, vol. 7, no. 1, 2018.
[15] D. Zhu, C. Cai, T. Yang, and X. Zhou, “A Machine Learning
REFERENCES Approach for Air Quality Prediction: Model Regularization and
Optimization,” Big Data and Cognitive Computing, vol. 2, no.
1, p. 5, Mar. 2018.
[1] Iskandar, Adha, M. B., Hendrawan, & Edward, I. Y. M. (2019). [16] A. Masih, “Machine learning algorithms in air quality
Design and Implementation of Web Application on Air Pollution modeling,” Global Journal of Environmental Science and
Monitoring System Using Wireless Sensor Network Based on Management, vol. 5, no. 4, pp. 515–534, 2019
HAPS. 2019 IEEE 5th International Conference on Wireless [17] Aditya C R, Chandana R Deshmukh, Nayana D K, Praveen
and Telematics (ICWT). Gandhi Vidyavastu;Detection and Prediction of Air Pollution
doi:10.1109/icwt47785.2019.897824110.1109/icwt47785.2019. using Machine Learning Models(IJETT)
8978241 [18] https://archive.ics.uci.edu/ml/datasets/Air+quality
[2] Ben-Aboud, Y., Ghogho, M., & Kobbane, A. (2020). A [19] David A. Freedman (2009). Statistical Models: Theory and
research-oriented low-cost air pollution monitoring IoT Practice. Cambridge University Press. p. 26. A simple
platform. 2020 International Wireless Communications and regression equation has on the right hand side an intercept and
Mobile Computing (IWCMC). an explanatory variable with a slope coefficient.
doi:10.1109/iwcmc48107.2020.9148176 [20] Rokach, Lior; Maimon, O. (2008). Data mining with decision
10.1109/iwcmc48107.2020.9148176 trees: theory and applications. World Scientific Pub Co Inc.
[3] Bashir Shaban, K., Kadri, A., & Rezk, E. (2016). Urban Air ISBN 978- 9812771711. [12] BreimanL
Pollution Monitoring System With Forecasting Models. IEEE (2001)."RandomForests". MachineLearning. 45 (1):32.
Sensors Journal, 16(8), 2598–2606. doi:10.1023/A:1010933404324
doi:10.1109/jsen.2016.2514378
[4] Okokpujie, Kennedy & Noma-Osaghae, Etinosa &
Odusami,Modupe & John, Samuel & Oluwatosin, Oluga.
(2018). A Smart Air Pollution Monitoring System. International
Journal of Civil Engineering and Technology. 9. 799-809.
[5] C R, Aditya & Deshmukh, Chandana & K, Nayana &
Gandhi, Praveen & astu, Vidyav. (2018). Detection and
Prediction of Air Pollution using Machine Learning Models.
International Journal of Engineering Trends and Technology.
[6] Arora, Himanshu & Solanki, Arun. (2020). Prediction of Air
Quality Index in Metro Cities using Time Series Forecasting
Models. Xi'an Jianzhu Keji Daxue Xuebao/Journal of Xi'an
University of Architecture & Technology. 12. 3052-3067.
10.37896/JXA 12.05/1721.
[7] Venkat Rao Pasupuleti , Uhasri , Pavan Kalyan , Srikanth and
Hari Kiran Reddy, Air Quality Prediction Of Data Log By
Machine Learning,2020,IEEE.
[8] Haotian Jing & Yingchun Wang, Research on Urban Air
Quality Prediction Based on Ensemble Learning of XGBoost,
2020, E3S Web of Conferences.
[9] Xiaosong Zhao , Rui Zhang, Jheng-Long Wu, Pei-Chann Chang
and Yuan Ze University, A Deep Recurrent Neural Network for
Air Quality Classification, 2018, Journal of Information Hiding
and Multimedia Signal Processing.
[10] Nicolás Mejía Martínez, Laura Melissa Montes, Ivan Mura and
Juan Felipe Franco, Machine Learning Techniques for PM10
Levels Forecast in Bogotá,2018,IEEE.
[11] Kennedy Okokpujie, Etinosa Noma-Osaghae, Odusami
Modupe, Samuel John, and Oluga Oluwatosin, “A SMART AIR
POLLUTION MONITORING SYSTEM,” International Journal
of Civil Engineering and Technology (IJCIET), vol. 9, no. 9, pp.
799–809, Sep. 2018.

ISSN : 2581-7175 ©IJSRED: All Rights are Reserved Page 159