Air Quality Index (AQI): Prediction and Optimization

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 02 | Feb 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1153
Air Quality Index (AQI): Prediction and Optimization
Dr. Jitendra Saturwar1, Baljit Thakur2, Nayan Siddhapura3, Dilip Sahani4
1Dept. Computer Engineering, Universal College of Engineering, Maharashtra, India
2,3,4Dept. Computer Engineering, Universal College of Engineering, Maharashtra, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - — People's health has been affected in the worst
way possible with the development of industrialization and
urbanization. Air pollution in many countries have become
more serious problem. The air quality has been continuously
concerned by environmentalist as well as the common people.
The objective of this project is to show the Air QualityIndexon
Azure maps to indicate deterioration of air quality so that
people should be aware of their environment. The Proposed
system can predict the Air quality of upcoming days and it can
also identify the places whose AQI index can possibly become
worse in upcoming time. Acting as a warning system for the
people to take appropriate measurestopreventitfrom further
deterioration. It is a web-based Platform so that people can
have easy access with an interactive map visualization. The
prediction model used in the project is Sessional
Autoregressive Integrated Moving Average (ARIMA), which is
widely used for forecasting methods.
Key Words: AQI, SARIMA, Forecasting, Map
visualization Azure, Air pollution.
1. INTRODUCTION
Air pollution causes serious harm to human and animal
health. Air pollution is caused by various factors that might
occur naturally or result from specific human activities. Air
Quality Index (AQI) is used to measure pollution present in
Air. The index reflects a scale that ranges from 0 to 500. The
higher the value of AQI, the greater the health risk
associated. An AQI value that's lower than 50 indicates little
to no risk, but a value that's 300 or higher means that the air
is hazardous to everyone.
Azure Maps is a collectionofgeospatial servicesthat
use fresh mapping data toprovidegeographic contextinweb
and mobile applications. The services include APIsformaps,
vehicle routing, weather, and geofencing. Azure Maps also
has a web SDK that you can use to display a map on a
webpage. The Azure Maps web SDK has a wide variety of
tools you can use to visualize spatial data on an interactive
map on a webpage. Before we get started adding a map to a
webpage, let's look at some of the capabilities of the SDK.
The Azure Maps web SDK has a wide variety of tools you can
use to visualize spatial data on an interactive map on a
webpage. The project is about getting the data from third
party API and displaying it on the webpage. The map will be
displayed and the different AQI values will be marked with
different colors.
In this project we will take a look at 24 Indian cities
air pollution levels over the years as well as forecast the air
pollution levels at the current rate of pollution for the entire
country. We will also try to explain how it varies season to
season. from the data given. We have usedstandardAQI -Air
quality Index, as our measure for the air pollution levels
We will be predicting values of AQI for India
through SARIMA model (Seasonal AutoregressiveIntegrated
Moving Average). It is one of the most widely used
forecasting methods for univariate time series data
forecasting. The dataset will be taken fromKaggletoprocess
the model and predict the AQI.
2. LITERATURE SURVEY
In this study Dong-Her Shih, Ting-Wei Wu, Wen-Xuan Liu
and Po-Yuan Shih will be completed on Azure cloud
computing platform using cloud services , according to
the characteristics of air quality monitoring data , with
Microsoft Azure Machine Learning service. The hourly
updated data provided by the government is used to
generate an alert / warning for deteriorating AQI. The data
used in this study is from January 2016 to May 2018 in
Taiwan and creates a prediction model of AQI pollutant
concentration. Using this data, the model predicts the AQI
warning for next six hours. It is expected that this study can
help public to avoid approaching the areas with serious air
pollution ,and to reduce the health hazards to
individuals.[1]
This study has been done by Huan Li ,Hong Fan and Fei
Yue Mao In this study A visual exploration method
was proposed to analyze air pollution data, which enables
rapid processing and multi - perspective exploration of
air pollution data to reveal spatio - temporal patterns and
basic relationships among multiple variables , The
outcome of this project suggests a strongcorrelationexisted
between pollutants and wind speed and Temporal
characteristics were found through visual analytics . From
October to March and timely from 6:00 p.m. to 4:00 a.m. the
concentrations are more serious. [2]
This paper presented by Appasani Geeta, Pasumarthi Sai
Ramya chenikala and Sravani M Ramesh This chapter
presents a study of analyzing the Air Quality Index using
individual methods and tools and to interpret the
significance of AQI in atmosphere. This empowers us to
know the perception of various creators and gives a clear

view of data interpretation and comparison between the
various tools.[3]
This paper published by Vineeta Ajit Bhat, Asha S Manek
and Pranay Mishra In thispaper,Theyhave proposeda novel
machine learning based system to estimate dense air
pollution using data collected from both from government
monitoring sites and wireless sensor . They choose three
regression models (Linear Regression, Random Forest
Regression and DecisionTreeRegression)andcompared the
estimation performances. They selected Random Forest
Regression (RFR) as their machine learning algorithm
because of low error index rate. They applied RFR algorithm
in their system for the optimal air pollutionlevelsestimation
performance.[4]
In this study Pranav Shrirama and Srinivash Malladi
proposed that Air pollution is a major problem for health,
sometimes healthy people also experience health impacts
from air pollution. Air pollution can cause respiratory
irritation, asthma, breathing problem during outdoor work.
Air polluted related health risks is depending on the current
health problem and impact is also varies from person to
person. The different health problem may cause by different
air particles present in the air also consider the factor of
pollutant type, the time duration of exposure, and air
particles concentration. It is essential to measure the air
quality parameter to identify the level of air pollution and
every country has its own standard measuring parameters
and guidelines of air quality index. [5]
In this Study Nimmathi Satheesh, N. Vallileka, S.Jai
Ganesh, R.SivaKumar, N.Muthu Lakshmi has presented
sharing the messages of Air Quality Index (AQI) in
Metropolitan areas. Air pollution is monitored using air
pollution sensors for outdoor environments and Cloud
technology is used to display to common people via private
cloud. Prototype development was aimed to create
awareness and people engagement in restoring the
environment back to its healthy state. By using this system,
the people can interact with the environment sensors in the
field of view to access and view latest and historic
environment measurements. [6]
3. PROPOSED SYSTEM
We have shown the detailed system Architecture in Fig 2.
AQI Prediction and optimization is a web app so that
everyone can access it on hand. It has two main elements i.e.
visualization of AQI data on azure maps which provides an
interactive platform for the people to view air quality index
of their region as well as the rest of the world. The second
phase deals with prediction of Future AQI using SARIMA
model. SARIMA is Seasonal ARIMA, or simply put, ARIMA
model with an additional seasonal component. As
mentioned, ARIMA is a statistical analysis model which uses
time-series data to better understand the data set and to
predict future trends. The SARIMAmodel equationlookslike
the following – Fig.1
Fig -1: Sarima Notation
The parameters for these types of models are p and
seasonal P indicate the number of AR terms (lags of
the stationary series), d and seasonal D indicate
differencing that must be done to stationary series, q and
seasonal Q indicate the number of MA terms (lags of the
forecast errors), lag indicates theseasonal lengthinthedata.
Fig. 2: System Architecture
The modules are:
3.1. Data Visualization.
For visualization process we will get the data through third
party API (i.e.waqi.info)[7]. The data which we will get from
the API will be in JSON format that will be converted to
GeoJSON before plotting . Then we will use Azure maps to
plot data on an interactive map . The AQI will be displayed
using bubble layer. Different AQI levels will be plotted using
different colors as per the standard chart.
Fig. 3: Data Visualization

3.2. AQI Prediction
The data taken from Kaggle. Data is then sent for
preprocessing where all the ambiguitiesareremoved. Notice
that the data is daily data. We will convert it into monthly
data for our ease by averaging a month’s data. After
preprocessing the data is trained using SARIMA (Seasonal
Autoregressive Integrated Moving Average) model which
produces forecasting data as output. The output will beused
as prediction for future possible AQI.
Fig. 4: Prediction model
3. Results and Discussion
The Pre-defined buckets of AQI levels are classified on the
basis of the following table:
Fig 5: Standard AQI
After converting Data from API into GeoJSON format the
data is being displayed on the map. User can navigate on the
map and check the AQI levels of their region as well as the
whole world and can understand as well as takeappropriate
measures to prevent it from deteriorating or can get
insightful information from the map visuals. Fig 6. displays
the map that users will be able to access.
Fig.6: Map Visualization
In preprocessing data, we have removed all the
ambiguities such as null values Fig. 7 represents
Ambiguous data and Fig.8 represents processed clean
data.
Fig.8: processed clean data
AQI calculation uses 7 measures: PM2.5(Particulate
Matter 2.5-micrometer), PM10 (Particulate Matter 10-
micrometer), SO2, NOx, NH3, CO and O3(ozone). For
calculating the AQI PM2.5, PM10, SO2, NOx and NH3 the
average value in last 24 hours is used and it must
contain at least 16 values. For CO and O3 the maximum
value in last eight hours is considered and then each
measure is converted into a Sub-Index based on pre-
defined groups.
The final AQI is the maximum sub-index with the
condition stating that at least one of PM2 and PM10
should be available and at least three out of the seven
pollutants should be available.
Fig. 7: Raw Dataset

Fig 9. Shows the graph of India's AQI over the years.
Fig.9 : India’s AQI
We start by running auto ARIMA to find out the
parameters of the model required. We can manually do it,
however, it is much easier for us let the notebook do the
work for us. Our next step is to forecast using thismodel into
the future. However, since we do not have information
regarding future values, we will split the data into a training
data and testing data and try to predict 1 year into the
future. We will use the years 2015-2020(till June) as our
train dataset and July-June the next year as our test dataset.
Further, we will predict into the year 2022.
We have provided our model with the training data and
the necessary parameters. The next step is to forecast the
next 12 months AQI values. To obtain the value of error we
have used root mean square error (RMSE) for comparison
between the models. We have got an RMSE value of
approximately 21, which is quite good and we can
approximately judge the scale of error by comparing with
the mean values of AQI which is 177 and hence the error is
approximately 1/9 of the actual values.
Next, we will proceed with forecasting model to predict
into the unknown data, i.e., year 2021 and2022.Thisposesa
problem, as if we predict including 2020 data, we are bound
to get an inaccurate prediction for next year simply due to
the fact that 2020 is an outlier. But if we remove 2020 from
our dataset and predict the values for 2021and2022. we are
left with wrong predictions for sure and considering that
covid-19 could have further lasting effects we will predict
poorly. We will choose to include 2020 as well for this
prediction. We could compare the values next year.
Fig 10: Forecasting model
Fig 10 shows the final model which displays the future
prediction based on the previous year’sdata.Theorangeline
in the graph represents the previous 5 years AQI values of
India. The blue line represents the predicted values of AQI
for the year 2021 and 2022. Here we can observe the
seasonal pattern that repeats itself year after another. This
signifies that there is a strong correlation between seasons
and AQI and which is repeated every year with slightdecline
in trend. Here we can see decrease in AQI values form the
year 2020 due to Covid-19 pandemic, which resulted in
decrease in pollution. The decrease in pollution is due to
Covid-19 restrictions which forced less movement of
unnecessary vehicles and major pollution causing factories
were closed.
4. CONCLUSION
Using Tracking of global air quality using azure maps and
the proposed system, it is very much possible for the
common civilians to access the AQI data and gain insights
about their region and any other geographical location. We
observed that the predicted values are close to our actual
values using SARIMA and hence is quite interesting how
looking at previous values gives us so much deeper insight
into future air pollution. However, there is a discrepancy at
the peak of the graph where our model has not been able to
predict with a high accuracy. To obtain the value of error we
will be using root mean square error(RMSE) for comparison
between the models. This project will create an awareness
among the common people and thus, people can take
necessary measures to reduce the health-related issues
caused by air pollution and identify the deteriorating
environment
REFERENCES
[1] Dong-Her Shih, Ting-Wei Wu, Wen-Xuan Liu, Po-Yuan
Shih, (2019), “An Azure ACES Early Warning System for
Air Quality Index Deteriorating” , MDPI
[2] Huan Li, Hong Fan, Feiyue Mao, “A Visualization
Approach to Air Pollution Data Exploration” ,(2016) ,
RESERCH GATE

[3] Appasani Geetha, Pasumarthi Sai Ramya, Chenikala
Sravani, M.Ramesh,“Real Time Air Quality Index from
Various Locations” ,(2020), IJERT
[4] Vineeta ,Ajit Bhat, Asha S Manek, Pranay Mishra,
“Machine Learning based Prediction System for
Detecting Air Pollution”, (2020), IJERT
[5] Pranav Shrirama, Srinivas Malladi, “ A Study and
Analysis of Air Quality Index and Related Health Impact
on Public Health” , (2020), ICICNIS.
[6] N. Vallileka, S.Jai Ganesh, Nimmathi
Satheesh, R.SivaKumar, N.Muthu
Lakshmi, “An Optimized Technique for
Sharing Air Quality Index Using Cloud
Technology” ,(2020) ,IJITEE.
[7] Microsoft docs: https://docs.microsoft.com/en-
us/learn/modules/azure-maps
[8] An Azure ACES Early Warning System for Air Quality
Index Deteriorating:
http://dx.doi.org/10.3390/ijerph16234679
[9] A Visualization Approach to Air Pollution Data
Exploration—A Case Study of Air Quality Index (PM2.5)
in Beijing, China
[10] World Health Organization Regional Office for Europe.
Monitoring Ambient Air Quality for Health Impact
Assessment; WHO Regional Office for Europe:
Copenhagen, Denmark, 1999.
[11] Air Quality index From Wikipedia, the freeencyclopedia
[12] Accuweather reference API

Air Quality Index (AQI): Prediction and Optimization

More Related Content

Similar to Air Quality Index (AQI): Prediction and Optimization (20)

More from IRJET Journal (20)

Recently uploaded (20)

Air Quality Index (AQI): Prediction and Optimization