Air_Quality_Index_Forecasting Prediction BP

Received 5 June 2023, accepted 21 June 2023, date of publication 30 June 2023, date of current version 7 July 2023.
Digital Object Identifier 10.1109/ACCESS.2023.3291146
Air Quality Index Forecasting via Genetic
Algorithm-Based Improved Extreme
Learning Machine
CHUNHAO LIU , (Graduate Student Member, IEEE), GUANGYUAN PAN , (Member, IEEE),
DONGMING SONG , AND HAO WEI
School of Automation and Electrical Engineering, Linyi University, Linyi 276000, China
Corresponding author: Dongming Song (david.dming@qq.com)
The research conducted in this study was supported by National Natural Science Foundation of China under Grant no. 62103177. This
funding played a crucial role in facilitating data collection, conducting experiments, and analyzing the results.
ABSTRACT Air quality has always been one of the most important environmental concerns for the general
public and society. Using machine learning algorithms for Air Quality Index (AQI) prediction is helpful
for the analysis of future air quality trends from a macro perspective. When conventionally using a single
machine learning model to predict air quality, it is challenging to achieve a good prediction outcome under
various AQI fluctuation trends. In order to effectively address this problem, a genetic algorithm-based
improved extreme learning machine (GA-KELM) prediction method is enhanced. First, a kernel method
is introduced to produce the kernel matrix which replaces the output matrix of the hidden layer. To address
the issue of the conventional limit learning machine where the number of hidden nodes and the random
generation of thresholds and weights lead to the degradation of the network learning ability, a genetic
algorithm is then used to optimize the number of hidden nodes and layers of the kernel limit learning machine.
The thresholds, the weights, and the root mean square error are used to define the fitness function. Finally, the
least squares method is applied to compute the output weights of the model. Genetic algorithms are able to
find the optimal solution in the search space and gradually improve the performance of the model through an
iterative optimization process. In order to verify the predictive ability of GA-KELM, based on the collected
basic data of long-term air quality forecast at a monitoring point in a city in China, the optimized kernel
extreme learning machine is applied to predict air quality (SO2, NO2, PM10, CO, O3, PM2.5 concentration
and AQI), with comparative experiments based CMAQ (Community Multiscale Air Quality), SVM (Support
Vector Machines) and DBN-BP (Deep Belief Networks with Back-Propagation). The results show that the
proposed model trains faster and makes more accurate predictions.
INDEX TERMS Time series, air quality forecasting, machine learning, extreme learning machine, genetic
algorithm.
I. INTRODUCTION
Air pollution is a prevalent environmental problem in the
twenty-first century. In light of the rapid industrialization and
urbanization, air pollution is getting worse, which greatly
affects our living environment and health [1]. Li et al. came to
the conclusion that outdoor physical activity poses numerous
health risks due to ambient air pollution in China. [2], [3].
According to the Chinese Ambient Air Quality Standards
The associate editor coordinating the review of this manuscript and
approving it for publication was Frederico Guimarães .
(GB3095-2012), there are six conventional air pollutants
used to measure air quality: sulfur dioxide (SO2), nitrogen
dioxide (NO2), particulate matter with a particle size less than
10 microns (PM10), particulate matter with a particle size less
than 2.5 microns (PM2.5), ozone (O3), and carbon monoxide
(CO) [4], [5], [6]. These pollutants have adverse effects on
human health. The International Energy Agency estimates
that air pollution causes 6.5 million premature deaths per
year, while long-term exposure to pollutants, such as fine
particles (e.g.,PM2.5) or traffic-related pollutants, is linked
to higher rates of lung cancer, coronary heart disease, and
67086
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ VOLUME 11, 2023

C. Liu et al.: AQI Forecasting via GA-KELM
other illnesses [7], [8]. Therefore, studies on air quality
prediction are particularly important and are considered a
key factor for environmental protection. In order to more
comprehensively assess the health effects of air pollution,
numerous air quality monitoring stations have been set up in
major cities. Air quality predictions can be made based on
the data collected from these stations. Air quality monitoring,
modeling, and accurate predictions are important for having
a clear understanding of future pollution levels and their
associated health risks.
Recently, the inherent property of machine learning
algorithms to automatically learn features at multiple levels of
abstraction has become increasingly important in providing
solutions to this challenging task [9], [10]. However, the
model only forecasts PM10 and SO2 levels, and it is
also challenging to obtain measurement values needed to
construct the dataset [11]. Wu Q. et al. proposed an
optimal-hybrid model for daily AQI prediction considering
air pollutant factors, with the model’s inputs being the six
atmospheric pollutants. However, neural networks typically
struggle with slow learning, a tendency to fall into local
minima, and a complex network training process. Based
on the generalized inverse matrix theory, Huang et al.
proposed an extreme learning machine (ELM) algorithm
with a feedforward neural network that includes a single
hidden layer, such that the problems of conventional neural
network algorithms are circumvented. The ELM algorithm
used to predict the AQI outperformed neural networks in
terms of parameter selection, training speed, and prediction
accuracy [12]. However, the parameters of the hidden layer
nodes and the number of nodes in the test hidden layer are
selected at random, which puts the prediction accuracy to a
great test.
In order to solve the aforementioned problems, we propose
to optimize the number of ELM hidden layer nodes,
thresholds, and weights, along with an improved genetic
algorithm (GA) that uses root mean square error (RMSE) as
the fitness function, to obtain the optimal network structure
for air quality prediction [14]. The number of hidden layer
nodes is updated by continuous coding discretization, the
input weights and hidden layer thresholds are updated by
continuous coding, and the update thresholds and weights
are selected with the number of updated layers to form a
hierarchical control structure [15]. The proposed GA-based
improved extreme learning machine (GA-KELM) algorithm
is applied to air quality prediction, and its performance is
compared with that of community multiscale air quality
modeling system (CMAQ), support vector regression (SVR),
and deep belief network-back propagation (DBN-BP). The
results show that the accuracy of the proposed GA-KELM
algorithm is reliable for air quality prediction [16].
In this study, an improved extreme learning machine
model based on a genetic algorithm is designed and
applied to AQI prediction. To verify the effectiveness of
the model, we conducted tests on three real-world datasets.
The results confirmed that the proposed method has superior
performance and outperforms some advanced methods cur-
rently in use. The main contributions of this paper are:
(1) modifying the ELM activation function or using the kernel
function to improve the prediction accuracy, (2) optimizing
the ELM using GA to improve the stability of the results and
further enhance the prediction accuracy, and (3) obtaining
the correlation analysis results of atmospheric environmental
quality prediction parameters by comprehensively consider-
ing each relevant factor in line with the actual situation. The
remainder of this paper is organized as follows. Section II
presents related work. Section III describes ELM and the
proposed GA-KELM, and illustrates the improvements using
the model. Section IV discusses experimental results where
GA-KELM is compared with several other methods in terms
of prediction results. The last section concludes the entire
work and presents directions for future research.
II. RELATED WORK
Air quality prediction has been extensively researched in
the literature [17]. In recent years, numerous researchers
have made significant contributions to the field by leveraging
quantitative studies and the latest techniques to identify
various air quality patterns and their underlying trends [18].
Existing work in this area relies on statistical methods and
shallow machine learning models to address the problem of
air quality prediction [19].
Agarwal and Sahu [20] conducted air quality prediction
studies by employing statistical models. Lary et al. [21] com-
bined remote sensing and meteorological data with ground-
based PM2.5 observations. Zheng et al. [22] proposed a hybrid
prediction method that combines a linear regression- based
temporal prediction method with an ANN-based spatial pre-
diction method for pollutant concentrations. Zheng et al. [23]
used a data-based approach for the next 48 hours of
PM2.5 prediction, implementing a prediction model based
on linear regression and neural network. They combined
meteorological data, weather forecast data, and air quality
data from monitoring stations. Rajput and Sharma [24] used
a multiple regression model to represent the changes in
air quality index (AQI), considering ambient temperature,
relative humidity, and barometric pressure as the main
parameters in the regression model for AQI calculation [25].
These classical methods and models all have the advantages
of simple algorithms, easy processing, and acceptable
prediction results. However, obtaining precise and specific air
quality prediction values remains challenging [26].
Elbaz et. al. [27] proposed a novel deep learning approach
that extracts high-level abstractions to capture the spatiotem-
poral characteristics of NEOM city in Saudi Arabia at
hourly and daily intervals. Campbell et al. [28] described
the development of FV3GFSv16 coupled with the ‘‘state-
of-the-art’’ CMAQ model version 5.3.1. Jin et al. [29]
proposed an interpretable variational Bayesian deep learning
model with self-filtering capability for PM2.5 prediction
information, which effectively improves prediction accuracy.
VOLUME 11, 2023 67087

Zhou et al. [30], [31], [32] proposed a method based on an
improved Grasshopper optimization algorithm to classify the
color difference of dyed fabrics using kernel extreme learning
machine. In this study, the classification of color differences
in dyed fabric images is performed using the kernel
limit learning machine, and the kernel function parameters
are optimized by the improved Grasshopper optimization
algorithm to achieve color difference classification of dyed
fabric images. Xue et al. [33] proposed a GA-based air
quality prediction model to optimize the parameters of the
weighted extreme learning machine (WELM). Despite the
progress made by the aforementioned methods, they also
exhibit limitations; their training efficiency is relatively low,
and deep learning algorithms are not yet fully mature. These
challenges present greater obstacles for the application of
deep learning, necessitating improvements to existing mod-
els, the development of new models, and the enhancement of
their predictive capabilities [34], [35].
The use of statistical or numerical forecasting techniques
is subject to several limitations. Neural networks are widely
used because of their unique associative abilities, memory,
and distinctive learning [36], [37]. Given the highly nonlinear
nature of AQI changes and the strong generalization and
nonlinear characterization abilities of neural networks, the
nuclear limit learning machine neural network model,
also known as kernel extreme learning machine (KELM),
is employed to investigate air quality prediction using a real
dataset. The weights and threshold values of KELM are
optimized using a genetic optimization algorithm [38].
III. METHODOLOGY
In this section, AQI is first introduced, the ELM and KELM
algorithms are presented next, and a new GA-KELM learning
method for AQI prediction is then proposed.
Air quality forecasting has been a key issue in early
warning and control of urban air pollution. Its goal is to
anticipate changes in the AQI value at observation points over
time. The observation period, which is decided by the ground-
based air-quality monitoring station, is usually set for one
hour.
Furthermore, a location’s air quality value is largely influ-
enced by the weather conditions prevailing at that location.
Air quality monitoring stations measure air temperature,
wind direction, atmospheric pressure, relative humidity, wind
speed and other meteorological parameters, as well as air
pollutant concentrations [39]. Air quality prediction is also
challenging due to the rapid changes in pollutant emission
and weather conditions. Numerous variables, such as wind
speed, temperature, humidity, and pollutants themselves, are
highly nonlinear, dynamic, and have inherent interdependen-
cies, making it more challenging to accurately predict air
quality at a specific time and place. Therefore, it is essential
to figure out how to deal with these factors and exploit
them from multivariate time-series data related to air quality.
A typical meteorological factor sequence diagram is shown
in FIGURE 1.
The AQI of each pollutant, or individual AQI (IAQI), is the
largest value of the day. AQI is calculated as in (1):
AQI = max {LAQI1, L AQI2, L AQI3, · · · , L AQIn} (1)
where LAQI1, L AQI2, L AQI3, · · · , L AQIn represent the
sub-index corresponding to each pollutant item. In this
study, the calculation of AQI involves only six pollu-
tants.Therefore, (1) can be expressed as:
AQI = max{IAQISO2 , IAQIN2 , IAQIO3 ,
IAQIPM10 , IAQIPM2.5 , IAQICO} (2)
To meet our research goals, this paper proposes
GA-KELM, a method for air quality prediction based on
an improved extreme learning machine, which in turn is
based on an improved GA [40]. Aiming at the problem of
network instability caused by the randomly generated input
layer weights and hidden layer thresholds of KELM, a GA is
used to optimize the KELM weights and thresholds, thereby
improving the model’s performance in terms of prediction
accuracy, which is the main objective of this algorithm.
In each iteration of the GA, a new offspring population
is generated by selection, crossover, and mutation, and the
individual with good fitness value is selected. The GA stops
iterating when the stopping criteria are satisfied. The GA is
used to determine the optimal weights and threshold values,
which overcomes the instability of KELM and reduces
prediction errors, thus resulting in a more reliable prediction
model and improved air quality prediction accuracy. Details
of the model will be discussed in the following sections.
A. EXTREME LEARNING MACHINE
ELM was first proposed by Huang. It is characterized by
its fast training and high training accuracy. Feedforward
neural networks are mainly based on the gradient descent
method [41]. Their main drawbacks are the slow training, the
tendency to fall into a local minimum point that cannot reach
the global optimum, and the high sensitivity to the learning
rate η (if the selected rate is improper, it might cause slow
convergence and the training process will thus take a long
time, or it becomes unstable). FIGURE 2 shows the network
structure of an ELM.
Consider N different samples (xi, ti) , i ∈ 1, 2, · · · N,
where xi denotes the input and ti represents the target, L
hidden layer neurons, and an activation function g(x).The
mathematical expression for ELM output is:
yi =
X
βjg wjxi + bj

(3)
where j ∈ 1, 2, · · · , L, wj are the weights of the input and
hidden layer neurons,bj are the thresholds of the hidden layer
neurons, βj is the weight matrix of neurons in the hidden and
output layers, and g wjxi + bj

is the output of the hidden
layer neurons.
67088 VOLUME 11, 2023

FIGURE 1. Sequence diagram of meteorological factors.
FIGURE 2. Schematic diagram of ELM structure.
In order to approximate the ideal output result, there must
be a set of wj, bJ and βj such that:
N
X
i=1
βjg wjxi + bj

= ti, i = 1, 2, · · · , N (4)
Like the single hidden layer neural network, (4) can be
expressed as:
Hβ = T (5)
where H is the output matrix of the hidden layer neurons and
T is the target output of the ELM network, i.e.,T = {ti}N
i=1,
so (5) can be transformed into the least squares problem of
solving for the weight matrix, namely:
∥Hβ̂ − T∥ = min
β
∥Hβ− | T∥ (6)
The output weight matrix β̂ can be obtained as:
β̂ = H+
T (7)
where H+ is the Moore-Penrose generalized inverse of H.
When the number of training observations is greater than the
number of hidden nodes, the expression β of the calculated
output weight is given by:
β̂ =

I
C
+ HT
H
−1
HT
T (8)
where I is the K-dimensional identity matrix and I
C is
the regularization parameter. When the number of training
samples is less than the number of hidden nodes, the output
weight β̂ expression is written as:
β̂ = HT

I
C
+ HHT
−1
T (9)
B. IMPROVED ELM BASED ON KERNEL METHOD
Inspired by SVM, the kernel method is introduced into
ELM, namely KELM. KELM maps linearly inseparable
patterns to a high-dimensional feature space to achieve
linear separability and improve the accuracy and robustness
of the model [42]. Therefore, it ensures that the network
produces good generalization and learns fast, and it avoids
the shortcomings of gradient descent training algorithms that
are prone to falling into local optima and necessitate a large
number of iterations.
ELM is a single hidden layer feedforward neural network,
and its learning objective (10) can be expressed as a matrix:
F(x) = h(x)β = Hβ = L (10)
Turning network training into a problem of solving a linear
system, β is determined by β = H ∗ L,where H∗ is a
generalized matrix of H.In order to enhance the stability of
the model, the regularization coefficient C and the identity
matrix I are introduced, and the output weights can be
VOLUME 11, 2023 67089

calculated by the least square’s solution in (11).
β = HT

HHT
+
I
c
−1
L (11)
In addition to the many advantages of the ELM algorithm,
KELM combines the kernel function to achieve linear
separability by non-linearly mapping the linearly inseparable
pattern to the high-dimensional feature space, which further
improves the judgmental accuracy. The kernel function is
introduced into ELM, and the kernel matrix is:
ELM = HHT
= h (xi) h xj

= K xi, xj

(12)
where xi and xj represent test input vector. Therefore,(10) can
be expressed as:
F(x) = [K (x, x1) ; . . . ; K (x, xn)]

I
C
+ ELM
−1
L
(13)
where x1, x2, · · · , xnis a given training sample, n is the
number of samples, and K(x1, xi) is the kernel function,
i = 1, 2, · · · , n. One of the widely used kernel functions is
the radial basis function (RBF). The RBF kernel can be
defined as:
K(x, y) = exp

−γ ∥x − y∥2

(14)
where y is the kernel parameter. Since the results of the
KELM model strictly depend on the regularization parameter
C and the choice of the kernel parameter y, these two
parameters need to be optimized effectively.
In conventional ELM, input weights and hidden layer
neurons are assigned at random, and output weights are then
obtained on the basis of input weights and hidden layer
neurons. Thus, human experiments are essentially important
in customary training, which often results in inaccurate
predictions [43]. To overcome this shortcoming, this paper
considers using a GA to optimize input weights and model
structure. In this section, as the input weights are continuous
variables and the model structure is discrete, the GA is
improved to handle the difference between continuous and
discrete variables of KELM in parallel optimization. The
improved GA-based KELM process is shown in FIGURE 3.
The steps for using a GA to optimize the parameters of
KELM are as follows:
Step 1: Data normalization. Since the observed data has
different properties, dimensions, and obvious differences in
numerical values, normalization is performed so that the
distribution range of the processed data is kept between
[0, 1], in order to avoid learning errors caused by the bias.
The normalization formula is given by:
X =
x − xmin
xmax − xmin
(15)
where x is the observed value, X is the input data of
the prediction model,xmin and xmax are the minimum and
maximum of the observed values, respectively [44].
FIGURE 3. GA-KELM algorithm flowchart.
FIGURE 4. Model architecture diagram.
Step 2: The input weights w and thresholds b of the hidden
layer of the ELM network are randomly generated as the
initial population of GA.
Step 3: The KELM model is trained using the training
samples, and the trained network is used to predict test
samples. The testing result is sent to the GA as objective
function.
Step 4: The improved GA is used to optimize weights and
model structure.
Step 5: Steps 3-4 are repeated until convergence is reached.
Step 6: The KELM is assigned the optimal hidden layer
input weights w and thresholds b for the final training and
testing.
C. OVERVIEW OF THE DEEP AIR QUALITY FORECASTING
FRAMEWORK
The flowchart of applying GA-KELM for AQI prediction is
shown in FIGURE 4.
a) Data collection and preparation. In GA-KELM, before
performing GA, the fitness value obtained during the
prediction process of the initial KELM is considered.
b) Prediction. In the experiment, two-thirds of the collected
data are used to train the KELM network, and the remaining
one-third is used as a test set.
c) Optimization. The GA is applied through the processes
of selection, crossover and mutation, until the optimal result
is obtained. The weights and structure parameters of KELM
are then updated using the values produced by the GA. This
process is repeated until the optimal result is achieved.
67090 VOLUME 11, 2023

All the spatiotemporal relationships that dominate the
input data are not entirely helpful in predicting the output.
Each feature has a level of association with the predicted
output. The goal of GA-KELM is to identify spatiotemporal
features and the degree of their association with the predicted
output learned through its weight matrix [45].
IV. EXPERIMENTAL RESULTS
In this section, experiments using real-world air quality
datasets are conducted to analyze and evaluate the perfor-
mance of the proposed model. The predictive performance of
the model is verified by comparing with other baseline deep
learning models.
A. DATASETS DESCRIPTION AND DATA PREPROCESSING
1) DATA COLLECTION AND DATA BASICS
For the study in this paper, we will use 3 real air
quality datasets. Our dataset is available on the following
link: https://github.com/chunhaoliu/AQI_Data.git. It includes
air quality data from July 2019 to July 2021, cover-
ing all seasons and different weather conditions. The
dataset contains multidimensional air pollutant monitoring
data, such as: SO2 µg/m3

, NO2 µg/m3

, PM10 µg/m3

,
PM2.5 mg/m3

, O3 µg/m3

and CO mg/m3

, hourly aver-
age concentration data and day-by-day detection data.
In the preprocessing stage, the data are initially examined
for missing value outliers. To ensure data completeness
and accuracy, linear interpolation and 3σ are applied for
processing. The original data are then sorted by time series
and divide it into a training set and a test set. This dataset
will be used to train and test our Improved Extreme Learning
Machine (ELM) model. With this approach, we can offer
crucial information regarding future air quality conditions to
the government and the public. This data-driven insight can
assist in the development and implementation of appropriate
environmental policies and measures.
2) DATA PREPROCESSING
When dealing with missing data, various methods such as
nearest neighbor filling, plural filling, median filling, and
mean filling have been widely discussed in the literature.
After conducting a thorough literature review and considering
the characteristics of the data at hand, the interpolation
method is selected as the preferred approach for filling the
missing data. This method identifies the closest sample point
as the fill value by searching sample points around the
missing data. The interpolation method offers advantages
such as reduced calculation, high processing efficiency,
high flexibility and ease of implementation. Moreover, this
approach has minimal impact on the correlation between the
data and ensures the integrity of the experimental data. The
interpolation effect is shown in FIGURE 5.
However, as observed in the aforementioned figure, the
presence of many outliers in the data is still evident. There-
fore, outlier processing becomes necessary. To eliminate the
FIGURE 5. Before data processing.
outliers in the observed data, statistical analysis methods are
usually used. To extract multiple observations and remove
outliers, statisitcal criteria such as the Lajda criterion and the
Grubbs criterion are commonly employed.
The Lajda criterion may not be highly accurate in removing
and identifying outliers when the number of measurements
is small (n10). This method follows a three-step process
for determining outliers. First, it assumes that the dataset
contains only chance errors. Second, it computes the standard
deviation of this set of data, and determines the interval
range by certain probability. As data within the range
of (µ-3σ, µ+3σ) account for 99.74% of the total data,
the probability of exceeding this range is less than 0.3%.
Therefore, the distribution characteristics of the vast majority
of the data align with the 3σ criterion, conforming to a normal
distribution.The standard deviation was calculated as follows:
Sg =
s
Pn
l Pi − P̄
2
n − 1
(16)
where Pi is the i th observation, Xg is the standard deviation
of the observation, P̄ is the average of n observations, and
n is the number of observations. For a suspicious data Xi,
if Vi = Xi − X̄ 3Sg, then X contains coarse deviation
and is rejected. The data are then processed using MATLAB
after using outliers, as shown in FIGURE 6, and they were
obviously smoothed.
B. EVALUATION METRICS
The GA-KELM framework is compared with three base-
line deep learning models: CMAQ, SVR, and DBN-BP.
CMAQ (Community Multiscale Air Quality) is a computer
model used to simulate and predict atmospheric pollutant
concentrations and air quality. SVR (Support Vector Regres-
sion) is a machine learning algorithm used to model and
predict regression problems. It is an extension based on the
support vector machine (SVM) algorithm designed to solve
VOLUME 11, 2023 67091

FIGURE 6. After data processing.
regression problems. DBN-BP (Deep Belief Network with
Backpropagation) is a neural network model that combines
deep confidence network (DBN) and backpropagation (BP)
algorithms. In the experiments, 70% of the data were used
for training and 30% for testing. The predictive performance
of the proposed model was assessed using several evaluation
metrics, such as RMSE, mean square error (MSE), and
correlation coefficient (R2).
RMSE quantifies the difference between predicted and
actual data. A low value of RMSE underlines a good forecast.
RMSE =
v
u
u
t
PN
i=1

Xi − X̂i
2
N
(17)
MSE is calculated as the squared difference between the
true value and the predicted value, which is then averaged.
It is defined over the range [0,+∞), where it is 0 when the
predicted value is exactly the same as the true value, and the
larger the error, the larger the value.
MSE =
1
n
n
X
i=1
ŷi − yi
2
(18)
R2 is a correlation coefficient that defines how well a
predicted value matches the original one. The higher the value
of R2, the better the predictive model. R2 is given by:
R2
= 1 −
P
Xi − X̂i
2
P
Xi − X̄
2
(19)
Air quality is measured using the AQI obtained by calcu-
lating the concentration of each pollutant being monitored.
The classification of pollutants in the AQI can help in
understanding the status of air pollutants. From the measured
concentration values, the AQI classification range for each
pollutant is strictly determined according to the Ambient
AQI, as shown in Table 2.The prediction performance of
GA-KELM is assessed by calculating the accuracy of the
predicted values in the actual categories.
C. CORRELATION WITHIN THE METEOROLOGICAL
PARAMETERS
This section examines the correlation between variables to
make more accurate predictions. When pollutant sources are
identified, pollutant concentrations are basically determined
by meteorological conditions without analyzing the changes
brought by the pollutant emissions to the AQI. Therefore,
a change in pollutant concentration inevitably causes a
change in AQI, and meteorological factors in turn determine
the pollutant concentration. Meteorological factors have a
significant role in AQI change. The main meteorological
factors that affect the AQI are climatic conditions, weather
conditions, temperature, wind, humidity, and precipitation.
Temperature is one of the most basic meteorological factors
among them. Higher temperatures increase the water vapor
content in the air. Increased wind speed accelerates air flow,
making it easier for pollutants to be transported to higher
altitudes, which in turn causes increased pollution at ground
level. Relative humidity is related to the moisture in the
air. Turbulence in the atmosphere has a diffusion effect on
pollutants, causing a decrease in the AQI in the region.
Furthermore, temperature and humidity significantly affect
the pollutant chemical reaction process. The wind direction
affects the diffusion of pollutants, which in turn causes
changes in pollutant concentrations and the AQI.
Therefore, for the impact of meteorological factors on AQI,
the correlation between the five variables of temperature,
humidity, air pressure, wind speed, wind direction and the
monitored concentrations of SO2, NO2, CO, PM10, PM2.5,
and O3 can be considered. Typically, to study the correla-
tion between two groups of variables x1, x2, x3, · · · , xp

,
y1, y2, y3, · · · yq

, the p ∗ q correlation coefficients between
the two sets of variables can be calculated, using an idea
similar to that of principal components. A certain linear
combination of each of the two sets of variables is identified
and the correlation between the linear combinations is
discussed. In order to have maximum correlation, it is
necessary to find the first pair of linear combinations in each
set of variables:

u1 = a11x1 + a21x2 + . . . ap1xp
v1 = b11y1 + b21y2 + . . . bq1yq
(20)
Similarly, the second pair of linear combinations with the
next largest correlation is found, so that they are respectively
uncorrelated with the first linear combination pair within this
group:

u2 = a12x1 + a22x2 + . . . ap2xp
v2 = b12y1 + b22y2 + . . . bq2yq
(21)
where u2 and v2 are correlated, and u1 is uncorrelated
with u2. In this way, r steps are carried out until the
correlation between the two groups of variables is extracted,
67092 VOLUME 11, 2023

TABLE 1. Correlation of various pollutants with five meteorological conditions.
and the r group of variables can be obtained, with
r ≤ min(p, q). To study the correlation between two sets of
random variables, the full correlation coefficient is used. The
correlation model selects temperature, humidity, barometric
pressure, wind speed, and wind direction as meteorological
factor indicators, assuming that the correspondence between
NO2 and SO2 (fuel combustion) is not taken into account.
SO2, NO2, CO, PM10, PM2.5, and O3 are selected as the
source indicators in the model, and all factors are selected
based on daily average values. The influence of each weather
factor on each pollutant is divided into positive and negative
correlations through separate statistical correlations.
Table 1 shows that the atmospheric water-holding capacity
is enhanced with temperature rise in typical times, the atmo-
spheric movement slows, the accumulation rate of pollutants
exceeds the diffusion rate, and pollutants are not easily
diffused, leading to an increase in pollutant concentration and
in AQI. The increase in humidity reflects the enhanced water
holding capacity of the atmosphere, which also makes the
accumulation rate of pollution higher than the diffusion rate,
thus causing the pollutant concentration to increase and the
AQI to rise. However, increased humidity can also be caused
by rainfall that can decrease pollutant concentrations, and
consequently, AQI is decreased. Pressure affects air quality
mainly as a result of seasonal changes rather than day-to-
day variations. For instance, low-pressure weather dominates
in winter and the atmosphere moves slowly, which is
unfavorable to pollutant dispersion. Wind speed is consistent
with wind direction. The analysis of the relationship between
meteorological factors and pollutants reveals that the total
amount of pollutant emissions varies greatly under different
meteorological conditions, and pollutant emissions decrease
with an increasing wind speed under the same meteorological
conditions. Conceptually, an increase in wind speed causes a
decrease in pollution concentration and a decrease in AQI at
the designated monitoring stations. However, the wind speed
increases in some direction whereas the AQI decreases at the
monitoring stations.
D. POLLUTANT CONCENTRATION FORECASTING RESULTS
ANALYSIS
In the experiment, a GA is used to optimize the parameters
of KELM. The fitness calculated by the GA is a weighted
combination of AQI and primary pollutant error. Data
from five previous hours (actual collected values including
pollutants and weather) are used as input. For instance,
assuming that the moment to be predicted is t, then data at
t − 1, t − 2, t − 3, t − 4, and t − 5 are taken as input. The
FIGURE 7. End-to-end architecture of GA-KELM.
FIGURE 8. Results of the change in fitness for GA optimization.
output is the pollutant concentration to be predicted at this
moment (i.e., at time t), as shown in FIGURE 7.
In the GA, the number of iterations is set to 100,
the population size is 10, the number of hidden layer
neurons is 30, the number of output layer neurons is
1 representing the predicted AQI, the number of training
iterations is 1000, the learning rate is 0.05, and the training
error target is set to 0.001. Additionally, the crossover
probability is varied in the range [0.6 1.0] in increments
of 0.05, and the variance probability is varied in the range
[0 1] with an increment of 0.01. RMSE is chosen as the
fitness parameter for the optimization problem. For each
parameter set, 20 independent replications are performed.
For each case, the RMSE values converge rapidly to a stable
point.
The parameters to be optimized, in this case the penalty
factor and the kernel parameters, are first set at random.
The data are divided into two parts: training and prediction
of the remaining days. The training samples and the set
parameters are used to train the model, and the test samples
are used to produce prediction results for AQI and major
pollutants. A weighted combination of the maximum relative
VOLUME 11, 2023 67093

FIGURE 9. Annual air quality forecast comparison (Dataset A).
FIGURE 10. Seasonal air quality forecast trends (Dataset B).
TABLE 2. Category range of air pollutants (technical regulation on ambient air quality index, HJ633-2012).
error of AQI and the average relative error of major pollutants
is used as our fitness function. Then the GA is used to
optimize the fitness function, and the penalty factor and
kernel parameter are used as the optimization parameters.
According to previous theory, the standard GA will complete
convergence at an early stage. The maximum evolutionary
generation is set to 40, and the kernel function selected
in KELM is the Gaussian RBF. The fitness curve of GA
optimization is shown in Fig.8. It can be observed that the
fitness keeps decreasing and finally reaches convergence.
At this point, the optimal penalty factor and kernel function
parameters can be obtained.
67094 VOLUME 11, 2023

FIGURE 11. Monthly air quality forecast accuracy (Dataset C).
TABLE 3. Comparison of RMSE of GA-KELM with other baseline
approaches.
TABLE 4. Comparison of MSE of GA-KELM with other baseline
approaches.
TABLE 5. Comparison of R2 of GA-KELM with other baseline approaches.
The obtained optimal parameters are then used to train
the limit learning machine. The actual values are compared
with the predictions of GA-KELM from July 2020 to July
2021; the red line represents the predicted values, and the
blue color represents the actual values. The results of three
different datasets A, B, and C are shown in FIGURE 9,
FIGURE 10, and FIGURE 11, respectively. As can be seen
from the figures, the training effect is relatively good.
In China, the National Air Quality Index sets breakpoints
for all air pollutants under 6 categories, namely good,
satisfactory, moderately polluted, poor, extremely poor, and
severe, as shown in Table 2. During the validation process,
the predicted values are grouped under each AQI category
based on the predicted value. The ratio of correct predictions
to total predictions is estimated to measure the accuracy of
the model in predicting the correct AQI category.
The performance of the model in predicting air pollutant
concentrations is measured based on the RMSE, MSE
and R2 values of all predicted pollutants. The results are
compared with baseline methods for predicting air pollutant
concentrations, such as CMAQ, SVR, and DBN-BP, with
meteorological parameters as auxiliary inputs. However, the
factors considered in model evaluation are different.
The proposed and baseline models were initially trained
using the training dataset, and from Table3-5, we infer that
GA-KELM air has smaller RMSE and MSE values and
higher R2 values compared to baseline methods for predicting
concentrations of various air pollutants.The reduced RMSE
and MSE values elucidate the reliability of the GA-KELM
air predictions, while the reduced RMSE and improved
R2 values infer the specificity of its mean predictions. The
superior performance of GA-KELM illustrates its efficiency
in accurately capturing spatiotemporal relationships and their
impact on predicted values.
V. CONCLUSION
The economic development achieved by the country through
rapid urbanization is polluting the environment in an alarming
way and putting people’s lives in danger. Therefore, a correct
analysis and accurate prediction of air quality remains a
primary condition to achieve the objective of sustainable
development. This paper focuses on the problem of prediction
model design, and investigates the problems related to the
optimization of the model parameters. A GA-KELM model
is designed, implemented, and tested. It is experimentally
proven to be more efficient than the classical shallow learning
and can effectively explore and learn the interdependence
of multivariate air quality correlation time series such
as temperature, humidity, wind speed, SO2, and PM10.
VOLUME 11, 2023 67095

Therefore, the GA-KELM model developed in this study can
be used to provide valuable support to vulnerable groups and
trigger early warning of adverse air quality events.
However, there are still areas for further investigation and
improvement. In recent years, numerous advanced algorithms
and optimization methods based on genetic algorithms and
population intelligence have emerged. Therefore, future
research should explore the underlying significance and value
of combinatorial intelligence optimization algorithms such as
the Limit Learning Machine. Additionally, we acknowledge
the need to address the issue of manually setting the number
of hidden layer nodes in the optimal Limit Learning Machine.
Although the Dynamic Extreme Learning Machine (DELM)
algorithm offers adaptive determination of hidden layer
nodes without human intervention, further work should be
dedicated to this aspect. Moreover, to enhance the accuracy
and validity of air quality measurement and assessment,
it is crucial to integrate pollutant emission factors and
meteorological factors into the evaluation system. This
integration will enable a more precise and comprehensive
evaluation of air quality.
In conclusion, our study highlights the significance of
the GA-KELM model in predicting air quality. We have
addressed the optimization challenges and demonstrated its
superiority over traditional methods. However, there is still
room for improvement and further research. Future studies
should delve into advanced optimization algorithms based on
genetic algorithms and population intelligence, explore the
potential of the Limit Learning Machine, and strive for adap-
tive determination of hidden layer nodes. Furthermore, the
integration of pollutant emission factors and meteorological
factors into the evaluation system will advance the accuracy
and reliability of air quality measurement and assessment.
REFERENCES
[1] X. Li, L. Jin, and H. Kan, ‘‘Air pollution: A global problem needs local
fixes,’’ Nature, vol. 570, no. 7762, pp. 437–439, Jun. 2019.
[2] Y. Han, J. C. K. Lam, and V. O. K. Li, ‘‘A Bayesian LSTM model to
evaluate the effects of air pollution control regulations in China,’’ in Proc.
IEEE Big Data Workshop (Big Data), Dec. 2018, pp. 4465–4468.
[3] L. Bai, J. Wang, X. Ma, and H. Lu, ‘‘Air pollution forecasts: An overview,’’
Int. J. Environ. Res. Public Health, vol. 15, no. 4, p. 780, 2018.
[4] Y. Ding and Y. Xue, ‘‘A deep learning approach to writer identification
using inertial sensor data of air-handwriting,’’ IEICE Trans. Inf. Syst.,
vol. E102-D, no. 10, pp. 2059–2063, 2019.
[5] S.-Q. Dotse, M. I. Petra, L. Dagar, and L. C. De Silva, ‘‘Application of
computational intelligence techniques to forecast daily PM10 exceedances
in Brunei Darussalam,’’ Atmos. Pollut. Res., vol. 9, no. 2, pp. 358–368,
Mar. 2018.
[6] M. Jia, A. Komeily, Y. Wang, and R. S. Srinivasan, ‘‘Adopting
Internet of Things for the development of smart buildings: A review of
enabling technologies and applications,’’ Automat. Construct., vol. 101,
pp. 111–126, May 2019.
[7] S. Abirami, P. Chitra, R. Madhumitha, and S. R. Kesavan, ‘‘Hybrid
spatio-temporal deep learning framework for particulate matter (PM2.5)
concentration forecasting,’’ in Proc. Int. Conf. Innov. Trends Inf. Technol.
(ICITIIT), Feb. 2020, pp. 1–6.
[8] Y. Cheng, S. Zhang, C. Huan, M. O. Oladokun, and Z. Lin, ‘‘Optimization
on fresh outdoor air ratio of air conditioning system with stratum
ventilation for both targeted indoor air quality and maximal energy
saving,’’ Building Environ., vol. 147, pp. 11–22, Jan. 2019.
[9] A. C. Cosma and R. Simha, ‘‘Machine learning method for real-time
non-invasive prediction of individual thermal preference in transient
conditions,’’ Building Environ., vol. 148, pp. 372–383, Jan. 2019.
[10] M. Bhowmik, K. Deb, A. Debnath, and B. Saha, ‘‘Mixed phase
Fe2O3/Mn3O4 magnetic nanocomposite for enhanced adsorption of
methyl orange dye: Neural network modeling and response surface
methodology optimization,’’ Appl. Organometallic Chem., vol. 32, no. 3,
p. e4186, Mar. 2018.
[11] V. Chaudhary, A. Deshbhratar, V. Kumar, and D. Paul, ‘‘Time series based
LSTM model to predict air pollutant’s concentration for prominent cities in
India,’’ in Proc. Int. Workshop Utility-Driven Mining (UDM), Aug. 2018,
pp. 1–9.
[12] M. Chen, J. Yang, L. Hu, M. S. Hossain, and G. Muhammad, ‘‘Urban
healthcare big data system based on crowdsourced and cloud-based air
quality indicators,’’ IEEE Commun. Mag., vol. 56, no. 11, pp. 14–20,
Nov. 2018.
[13] R. Chen, X. Wang, W. Zhang, X. Zhu, A. Li, and C. Yang, ‘‘A hybrid
CNN-LSTM model for typhoon formation forecasting,’’ GeoInformatica,
vol. 23, no. 3, pp. 375–396, Jul. 2019.
[14] S. Du, T. Li, Y. Yang, and S. Horng, ‘‘Deep air quality forecasting using
hybrid deep learning framework,’’ IEEE Trans. Knowl. Data Eng., vol. 33,
no. 6, pp. 2412–2424, Jun. 2021.
[15] R. Feng, H.-J. Zheng, H. Gao, A.-R. Zhang, C. Huang, J.-X. Zhang, K. Luo,
and J.-R. Fan, ‘‘Recurrent neural network and random forest for analysis
and accurate forecast of atmospheric pollutants: A case study in Hangzhou,
China,’’ J. Cleaner Prod., vol. 231, pp. 1005–1015, Sep. 2019.
[16] B. S. Freeman, G. Taylor, B. Gharabaghi, and J. Thé, ‘‘Forecasting air
quality time series using deep learning,’’ J. Air Waste Manage. Assoc.,
vol. 68, no. 8, pp. 866–886, Aug. 2018.
[17] S. Mahajan, H.-M. Liu, T.-C. Tsai, and L.-J. Chen, ‘‘Improving the
accuracy and efficiency of PM2.5 forecast service using cluster-based
hybrid neural network model,’’ IEEE Access, vol. 6, pp. 19193–19204,
2018.
[18] J. Jin, J. Gubbi, S. Marusic, and M. Palaniswami, ‘‘An information
framework for creating a smart city through Internet of Things,’’ IEEE
Internet Things J., vol. 1, no. 2, pp. 112–121, Apr. 2014.
[19] A. Grover, A. Kapoor, and E. Horvitz, ‘‘A deep hybrid model for
weather forecasting,’’ in Proc. 21st ACM SIGKDD Int. Conf. Knowl.
DiscoveryData Mining, Aug. 2015, p. 379–386.
[20] A. Agarwal and M. Sahu, ‘‘Forecasting PM2.5 concentrations using
statistical modeling for Bengaluru and Delhi regions,’’ Environ. Monit.
Assessment, vol. 195, p. 502, Mar. 2023.
[21] D. J. Lary, T. Lary, and B. Sattler, ‘‘Using machine learning to estimate
global PM2.5 for environmental health studies,’’ Environ. Health Insights,
vol. 9, no. 1, pp. 41–52, 2015.
[22] Y. Zheng, X. Yi, M. Li, R. Li, Z. Shan, E. Chang, and T. Li,
‘‘Forecasting fine-grained air quality based on big data,’’ in Proc. 21th
ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, New York, NY,
USA, Aug. 2015, pp. 2267–2276.
[23] Y. Zheng, L. Capra, O. Wolfson, and H. Yang, ‘‘Urban computing:
Concepts, methodologies, and applications,’’ ACM Trans. Intell. Syst.
Technol., vol. 5, no. 3, p. 38, Sep. 2014.
[24] T. S. Rajput and N. Sharma, ‘‘Multivariate regression analysis of air quality
index for Hyderabad city: Forecasting model with hourly frequency,’’ Int.
J. Appl. Res., vol. 3, no. 8, pp. 443–447, 2017.
[25] Z. Kang and Z. Qu, ‘‘Application of BP neural network optimized by
genetic simulated annealing algorithm to prediction of air quality index
in Lanzhou,’’ in Proc. IEEE Comput. Intell. Appl. (ICCIA), Sep. 2017,
pp. 155–160, doi: 10.1109/CIAPP.2017.8167199.
[26] B. Liu, S. Yan, J. Li, G. Qu, Y. Li, J. Lang, and R. Gu, ‘‘A sequence-to-
sequence air quality predictor based on the n-step recurrent prediction,’’
IEEE Access, vol. 7, pp. 43331–43345, 2019.
[27] K. Elbaz, I. Hoteit, W. M. Shaban, and S.-L. Shen, ‘‘Spatiotemporal air
quality forecasting and health risk assessment over smart city of NEOM,’’
Chemosphere, vol. 313, Feb. 2022, Art. no. 137636.
[28] P. C. Campbell, Y. Tang, P. Lee, B. Baker, D. Tong, R. Saylor,
A. Stein, J. Huang, H.-C. Huang, E. Strobach, J. McQueen, L. Pan,
I. Stajner, J. Sims, J. Tirado-Delgado, Y. Jung, F. Yang, T. L. Spero, and
R. C. Gilliam, ‘‘Development and evaluation of an advanced national air
quality forecasting capability using the NOAA Global Forecast System
version 16,’’ Geosci. Model Develop., vol. 15, no. 8, pp. 3281–3313,
Apr. 2022.
67096 VOLUME 11, 2023

[29] X.-B. Jin, Z.-Y. Wang, W.-T. Gong, J.-L. Kong, Y.-T. Bai, T.-L. Su,
H.-J. Ma, and P. Chakrabarti, ‘‘Variational Bayesian network with infor-
mation interpretability filtering for air quality forecasting,’’ Mathematics,
vol. 11, no. 4, p. 837, Feb. 2023.
[30] J. Li, W. Shi, and D. Yang, ‘‘Color difference classification of dyed fabrics
via a kernel extreme learning machine based on an improved grasshopper
optimization algorithm,’’ Color Res. Appl., vol. 46, no. 2, pp. 388–401,
Apr. 2021.
[31] Z. Zhou, Y. Hu, J. Ji, Y. Wang, Z. Zhu, D. Yang, and J. Chen, ‘‘Robot
manipulator visual servoing via Kalman filter-optimized extreme learning
machine and fuzzy logic,’’ KSII Trans. Internet Inf. Syst., vol. 16, no. 8,
pp. 2529–2551, 2022.
[32] Z. Zhou, W. Deng, Z. Zhu, Y. Wang, J. Du, and X. Liu, ‘‘Fabric defect
detection based on feature fusion of a convolutional neural network and
optimized extreme learning machine,’’ Textile Res. J., vol. 92, nos. 7–8,
pp. 1161–1182, Apr. 2022.
[33] Y. Xue, Q. Wu, and H. Sun, ‘‘Air quality prediction model based on
genetic algorithm and weighted extreme learning machine,’’ Proc. SPIE,
vol. 12294, Oct. 2022, Art. no. 1229454.
[34] X. Liang, G. Wang, M. R. Min, Y. Qi, and Z. Han, ‘‘A deep spatio-temporal
fuzzy neural network for passenger demand prediction,’’ in Proc. SIAM Int.
Conf. Data Mining (SDM), 2019, pp. 100–108.
[35] Y. Liu, C. Lyu, A. Khadka, W. Zhang, and Z. Liu, ‘‘Spatio-temporal
ensemble method for car-hailing demand prediction,’’ IEEE Trans. Intell.
Transp. Syst., vol. 21, no. 12, pp. 5328–5333, Dec. 2020.
[36] J. Ma, Y. Ding, J. C. P. Cheng, F. Jiang, and Z. Wan, ‘‘A temporal-spatial
interpolation and extrapolation method based on geographic long short-
term memory neural network for PM2.5,’’ J. Cleaner Prod., vol. 237,
Nov. 2019, Art. no. 117729.
[37] J. Ma, Y. Ding, V. J. L. Gan, C. Lin, and Z. Wan, ‘‘Spatiotemporal
prediction of PM2.5 concentrations at different time granularities using
IDW-BLSTM,’’ IEEE Access, vol. 7, pp. 107897–107907, 2019.
[38] S. Maji, S. Ghosh, and S. Ahmed, ‘‘Association of air quality with
respiratory and cardiovascular morbidity rate in Delhi, India,’’ Int.
J. Environ. Health Res., vol. 28, no. 5, pp. 471–490, Sep. 2018.
[39] M. Wang, H. Chen, B. Yang, X. Zhao, L. Hu, Z. Cai, H. Huang, and
C. Tong, ‘‘Toward an optimal kernel extreme learning machine using a
chaotic moth-flame optimization strategy with applications in medical
diagnoses,’’ Neurocomputing, vol. 267, pp. 69–84, Dec. 2017.
[40] M. A. E. Aziz, A. A. Ewees, and A. E. Hassanien, ‘‘Whale optimization
algorithm and moth-flame optimization for multilevel thresholding image
segmentation,’’ Expert Syst. Appl., vol. 83, pp. 242–256, Oct. 2017.
[41] Y. Yang, Y. Wang, and X. Yuan, ‘‘Parallel chaos search based incremental
extreme learning machine,’’ Neural Process. Lett., vol. 37, no. 3,
pp. 277–301, Jun. 2013.
[42] F. Jiang, J. He, and Z. Zeng, ‘‘Pigeon-inspired optimization and
extreme learning machine via wavelet packet analysis for predicting bulk
commodity futures prices,’’ Sci. China Inf. Sci., vol. 62, no. 7, pp. 1–19,
Jul. 2019.
[43] Q. Li, H. Chen, H. Huang, X. Zhao, Z. Cai, C. Tong, W. Liu, and X. Tian,
‘‘An enhanced grey wolf optimization based feature selection wrapped
kernel extreme learning machine for medical diagnosis,’’ Comput. Math.
Methods Med., vol. 2017, Jan. 2017, Art. no. 9512741.
[44] G.-B. Huang, D. H. Wang, and Y. Lan, ‘‘Extreme learning machines:
A survey,’’ Int. J. Mach. Learn. Cybern., vol. 2, no. 2, pp. 107–122,
Jun. 2011.
[45] G. Huang, H. Zhou, X. Ding, and R. Zhang, ‘‘Extreme learning machine
for regression and multiclass classification,’’ IEEE Trans. Syst., Man,
Cybern. B, Cybern., vol. 42, no. 2, pp. 513–529, Apr. 2012.
CHUNHAO LIU (Graduate Student Member,
IEEE) was born in Linyi, Shandong, China,
in 1997. He received the B.S. degree in automation
from Linyi University, China, in 2019, where
he is currently pursuing the Graduate degree in
electronics and information technology with the
School of Automation and Electrical Engineering.
His current research interests include machine
learning and deep learning, data mining, and
control systems.
GUANGYUAN PAN (Member, IEEE) was born
in Linyi, Shandong, China, in 1987. He received
the B.S. degree in automation from the Qingdao
University of Science and Technology, China,
in 2009, and the M.S. and Ph.D. degrees in
control science and engineering from the Beijing
University of Technology (Dr. Junfei Qiao’s team),
Beijing, China, in 2012 and 2016, respectively.
From 2013 to 2014, he was a joint-training Ph.D.
student with the University of Victoria (Dr. Nikitas
Dimopoulos’ team), BC, Canada. He was a Postdoctoral Researcher with
the Department of Civil and Environmental Engineering, University of
Waterloo, ON, Canada, from 2016 to 2019. He is currently an Associate
Professor with the Department of Automation and Electrical Engineering,
Linyi University, China. His current research interests include machine
learning and deep learning, data mining and signal processing, transportation
safety, and intelligent systems.
DONGMING SONG was born in Yantai,
Shandong, China, in 1976. He received the
bachelor’s degree in mechanical design and
manufacturing from the China University of
Petroleum (East China), in 1999, and the master’s
degree in mechanical and electronic engineering
and the Ph.D. degree in computer application
technology from the Nanjing University of Science
and Technology, in 2004 and 2009, respectively.
He is currently an Associate Professor with the
Department of Automation and Electrical Engineering, Linyi University,
China. His current research interests include energy internet, energy big data,
electrical, and control systems.
HAO WEI was born in Linyi, Shandong, China,
in 1989. He received the bachelor’s degree in
computer science from Imperial College London,
U.K., in 2011, and the master’s degree in software
engineering and the Ph.D. degree in computer
science from the University of York, U.K., in
2013 and 2018, respectively. He is currently a
Lecturer with the Department of Automation and
Electrical Engineering, Linyi University, China.
His current research interests include swarm
robotics, genetic algorithm, and system development related to artificial
intelligence and big data.
VOLUME 11, 2023 67097

Air_Quality_Index_Forecasting Prediction BP

More Related Content

Similar to Air_Quality_Index_Forecasting Prediction BP (20)

Recently uploaded (20)

Air_Quality_Index_Forecasting Prediction BP