Abstract:
Clean air is the need of every living thing. Air pollution has become one of the
most alarming problem of today. It has become so intense that many people lost their
lives because of this problem. Pollution has a direct impact on human health. Some of the
common air pollutants such as SO2, CO, NO2 and PM2.5, PM10 influence human lives.
Man has undergone this effect in many countries. Diseases such as respiratory diseases,
asthma, chronic diseases, cardiovascular and pulmonary cancer. Air pollution is
increasing day by day. Increase in daily population also increases the pollution problems.
The increase in smog and darkness also increased the number of road accidents due to
poor visibility with dense smog. The concentration of PM2.5 in the environment was
found in Lahore at 130 µg/m3
, the normal value of which is 10 µg/m3
according to World
Health Organization (WHO). For this reason, serious measures are urgently needed to
solve the problem. It is necessary to foresee the relationship between air pollutants and
the environment, which will therefore help to control the excessive increase in air
pollution. In this work, the concentration of PM2.5 in the environment of Guangzhou
City of Guangdong province of China is predicted using Least Square Support Vector
Machine (LSSVM). Firstly the data of Gaungzau for the input parameters that influence
the concentration of PM2.5 are collected. Hourly air pollution data from January 2018 to
September 2020 was collected at the Guangzhou Environmental Monitoring Center.
There are 11 aerial surveillance stations in Guangzhou. Daily concentrations of PM2.5
and other air pollutants (SO2, NO2 and O3) were observed and the authorities regularly
monitored these stations. Emissions from traffic, industrial resources, buildings, or
residential emissions from coal, waste or oil combustion around the station are
considered to represent the overall Guangzhou air pollution statistics. Missing values in
the required data were imputed with the help of Principle Component Analysis (PCA).
LSSVM predicted the required quality parameter which is PM2.5 and evolutionary
optimization algorithms were applied including the Teaching-Learning Self Study-x
Optimization (TLSO), Particle Swarm Optimization (PSO) and Ant colony Optimization
(ACO). All of these techniques are applied on MATLAB. MATLAB is a software with
the best computational skills and is well known for its matrix computational skills.
Utilizing the data of routine observations, concentrations of air pollutants in the next day
was predicted. The result shows that the model developed by LSSVM optimized by PSO
in comparison with the LSSVM optimized with TLSO and ACO outperformed in
predicting PM2.5. Based on the predictions, people can adapt their activities and prepare
in advance to avoid air pollutants, and therefore can mitigate the effects on health and
economy.
