http://www.jhsci.ba
Pokkuluri Kiran Sree and S.S.S.N.Usha Devi Nedunuri. Journal of Health Sciences 2020;10(1):34-38
Journal of Health Sciences
RESEARCH ARTICLE
Open Access
A novel cellular automata classifier for COVID-19
trend prediction
Pokkuluri Kiran Sree1*, S. S. S. N. Usha Devi Nedunuri2
1
Department of Computer Science and Engineering, Shri Vishnu Engineering College for Women, West Godavari District,
Andhra Pradesh, India, 2Department of Computer Science and Engineering, University College of Engineering, JNTU
Kakinada, Andhra Pradesh, India
ABSTRACT
Introduction: China has witnessed a new virus Corona,which is named COVID-19. It has become the
world’s most concern as this virus has spread over the worldat a higher speed;the world has witnessed
more than one lakh cases and one thousand deaths in a span of few days.
Methods: We have developed a preliminary classifier with non-linear hybrid cellular automata, which is
trained and tested to predict the effect of COVID-19 in terms of deaths, the number of people affected,
the number of people being could be recovered, etc. This indirectly predicts the trend of this epidemic in
India. We have collected the datasets from Kaggle and other standard websites.
Results: The proposed classifier, hybrid non-linear cellular automata (HNLCA), was trained with 23,078
datasets and tested with 6785 datasets. HNLCA is compared with conventional methods of long shortterm memory, AdaBoost, support vector machine, regression, and SVR and has reported an accuracy of
78.8%, which is better compared with the cited literature. This classifier can also predict the rate at which
this virus spreads, transmission within the boundary, and of the boundary, etc.
Keywords: COVID-19 prediction; cellular automata; non-linear CA
INTRODUCTION
The coronavirus sickness 2019 (COVID-19)
flare-up starting in Wuhan, Hubei area, and China,
agreed with Chunyun, the time of mass movement
for the yearly spring festival. To contain its spread,
China embraced uncommon across the country
intercessions on January 23, 2020. These strategies included huge scope isolate, severe controls
on movement, and broad observation of suspected
*Corresponding author: Pokkuluri Kiran Sree, Department of
CSE, Shri Vishnu Engineering College for Women, West Godavari
District, Andhra Pradesh, India. E-mail: drkiransree@gmail.com
Submitted: 26 March 2020/Accepted: 29 March 2020
DOI: https://doi.org/10.17532/jhsci.2020.907
81,9(56,7<2)6$5$-(92
)$&8/7<2)+($/7+678',(6
cases. In any case, it is obscure whether these strategies have affected the pestilence. We looked to show
how these control measures affected the regulation
of the pandemic. Cellular automata area technique that operates basically with fuzzy rules. We
have done a feasibility study of what type of classifier is used to address this problem as the input is
very dynamic, and the prediction rate is also low.
We have applied non-linear cellular automata with
hybrid rules to classify the data and predict the people infected probable deaths and probable recoveries, etc. Cellular automata are natural classifier that
can classify dynamic data. Initially, we thought of
applying a deep learning model to predict the effect
of COVID-19, later we understood applying CA
© 2020 Pokkuluri Kiran Sree and S. S. S. N. Usha Devi Nedunuri; licensee University of
Sarajevo - Faculty of Health Studies. This is an Open Access article distributed under the
terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/
by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Pokkuluri Kiran Sree and S.S.S.N.Usha Devi Nedunuri. Journal of Health Sciences 2020;10(1):34-38
itself will give a better classifier when the dynamic
rules are used.
TABLE 1. Rules with neighborhood
Neighbors 111 110 101 100 011 010 001 000 Rule
R-51
0
0
1
1
0
0
1
1
51
R-238
1
1
1
0
0
1
1
1 238
Yang et al. (1) have studied the impact of COVID
on China based on many statistical parameters.
They have used a susceptible-exposed-infectious-removed model to predict the curve of propagation,
which is very preliminary work. This work is not
adoptable and about a limited area.Huet al.(2) have
proposed an AI technique to forecast the moments
of COVID-19.The authors have utilized the idle factors in the autoencoder and bunching calculations
to gather the areas/urban communities for exploring
the transmission structure. Yanet al. (3) have done
basic research on the moments of the COVID virus.
Zheet al.(4) have studied the primary reasons for
this epidemic. We have done an extensive literature
survey(5–8),on what type of classifier can be applied
to this problem. After this study, we understood that
cellular automata could be applied to this problem.
The rules selected are non-linear and hybrid ones.
TABLE 2. Rules without complementation
1.
2.
3.
4.
5.
METHODS
(1)
The transition from one state to another state is
defined by the rules, as described in the equation.
Algorithm (HNLCA Algorithm)
Input: Nine parameters collected and DNA
sequence
Output: Prediction(rate of death, infected,
andcurable)
Step 1: Embedding the sequence and preprocess the
parameters.
Step 2: Use the rules as Stated 1,2, and 3.
Step 3: Feature identification.
Step 4: Construct the HNLCA classifier as represented in section 3.3.
(2)
Non complemented 238: qi(t+1) = qi(t)+qi+1(t) (2)
3.1 Example: The classifier working will be explained
with the sample input(9).
Input: AAAGGGGACGTTTA
Embedding
layer:
0.8………………………….
qi-1+qi+1
qi
q i-1
qi+1
0
Design of hybrid non-linear cellular automata
(HNLCA)
Wolfram (8) has defined, the rule number can be
the decimal number equivalent of the next state as
shown in Tables 1 and 2. In this context, when we
consider three neighborhoods and two states for cellular automata, we can give 256 different next states.
Complemented Rule 51: qi(t+1) = qi(t)
250
204
240
170
0
Output: Basin with characteristics (the basin in the
example is State:4).
We have considered various parameters for predicting the spread of the virus listed below.
1. Vu: Vulnerable people in the region number
2. InOu: Inward and outward movement of the
people
3. Ti: Transmission rate of possible infections
4. Te: Transmission rate of possible exposure
5. Vo: Volatility of transfer per person wrt today
6. P: Total population in the regions
7. D: Theprobability of death
8. R: The probability to recover
9. E: The probability to get effected
As we know, cellular automata are a versatile classifier that can process dynamic data. The general rules
of CA can be represented like equation 1.
T(n+1)=T(n-1)+T(n)+T(n+1)
http://www.jhsci.ba
0.6,0.2,0.1,0.7,
Transitions: Processing by CA(Rule 204, 170, 250)
State1: 0.6 0.2 0.1
State 2: 0.6 0.2 0.1
State 3: 0.6 0.1 0.1
State 4:0.6 0.1 0.1
Real-time processing of HNLCA
The input from (9) was processed that looks like the
structure. The output of the classifier is 0.4, 0.4, and
0.8, which shows that the DNA sequence processed
35
http://www.jhsci.ba
Pokkuluri Kiran Sree and S.S.S.N.Usha Devi Nedunuri. Journal of Health Sciences 2020;10(1):34-38
has a chance of 0.4 infected, 0.4 death, and 0.8
chance of recovery in Figure 1.
observed that the error rate is very less in HNLCA,
which is 11.8% better compared to SVR. HNLCA
reports an accuracy of 78.8, which is better than
long short-term memory.
Figure 3 represents how HNLCA processed has
different parameters and can predict the rate of
changes in death, infected, and recovery rate.
HNLCA was applied to a small village Chodavaram
RESULTS
The datasets are collected from Kaggle, which consists of state, region, latitude, longitude, date, confirmed, deaths, and recovered, as represented in
(10-13). Figure 2 shows the comparison of the proposed classifier with the standard classifiers. We have
FIGURE 1. Real-time processing of hybrid non-linear cellular automata.
FIGURE 2. Performance of HNLCA and comparisons (HNLCA: Hybrid non-linear cellular automata, LSTM: Long short-term memory,
AdaBoost: Adaptable boosting, SVM: Support vector machine, REG: Regression, SVR: Support vector regression).
36
Pokkuluri Kiran Sree and S.S.S.N.Usha Devi Nedunuri. Journal of Health Sciences 2020;10(1):34-38
http://www.jhsci.ba
FIGURE 3. COVID-19 predictions with different parameters. (Vu: Vulnerable people in the region number, InOu: Inward and outward
movement of the people,Ti: Transmission rate of possible infections,Te: Transmission rate of possible exposure, Vo: Volatility of transfer
per person wrt today, P: Total population in the regions, D: The probability to death, R: The probability to recover,E: The probability to get
affected).
FIGURE 4. COVID-19 cases prediction in India.
in Andhra Pradesh, where the number of people
residing is1799, vulnerable people in the region
number are 911, inward and outward movement of
the people are 162, the transmission rate of possible
infections 154, and the transmission rate of possible
exposure 406. The classifier predicts that the volatility of transfer per person today probability is 0.022,
the probability to death 0.0055, The probability to
recover 0.0072, and The probability to get effected
is very less. Figure 4 predicts the total number of
cases in India every day. This graph shows that the
number of people affected is gradually increasing
every day, and on March 24 represents the maximum number of cases reported. Figure 5 shows the
number of deaths predicted for the past 15 days. The
number of deaths from February 15 to March 10
37
http://www.jhsci.ba
Pokkuluri Kiran Sree and S.S.S.N.Usha Devi Nedunuri. Journal of Health Sciences 2020;10(1):34-38
FIGURE 5. COVID-19 death prediction in India.
3. Bai Y, Yao L, Wei T, Tian F, Jin DY, Chen L, et al. Presumed asymptomatic
carrier transmission of COVID-19. JAMA 2020; ahead of print,
was zero, and after that, there is consistent growth
in the number of deaths.
https://doi.org/10.1001/jama.2020.2565
4. Xu Z, Shi L, Wang Y, Zhang J, Huang L, Zhang C, et al. Pathological
findings of COVID-19 associated with acute respiratory distress syndrome. Lancet Respir Med 2020; online first, https://doi.org/10.1016/
S2213-2600(20)30076-X
CONCLUSION
We have successfully developed as a novel classifier
HNLCA, which can process real-time datasets to
predict the number of deaths and cases in the near
future. This can also predict the number of people
that can recover, the rate at which this virus spreads,
transmission within the boundary and of the
boundary, etc. The classifier is thoroughly trained
and tested. The developed classifier is adaptable to
various dynamics inputs and can process data of
different lengths. Even though this is a preliminary
work to predict trends of COVID, we have developed the first and novel classifiers with dynamic
rules with cellular automata. This work can be future
extended with incorporating fuzzy sets augmented
with CNN, which will give the best performance.
The average accuracy of 78.8% is reported, which is
considerable at this moment.
5. Kiran SP, Babu IR. Identification of protein coding regions in genomic
DNA using unsupervised FMACA based pattern classifier. IJCSNS
2008;8(1):305.
6. Kiran SP, Babu IR, Devi U. Investigating an artificial immune system to
strengthen protein structure prediction and protein coding region identification using the cellular automata classifier. Int J Bioinform Res Appl
2009;5(6):647-62.
https://doi.org/10.1504/ijbra.2009.029044.
7. Pokkuluri KS, Babu IR. AIX-MACA-Y multiple attractor cellular automata
based clonal classifier for promoter and protein coding region prediction.
J BioinformIntell Control 2014;3(1):23-30.
https://doi.org/10.1166/jbic.2014.1071.
8. Wolfram S. Universality and complexity in cellular automata. Physica D
1984;10(1-2):1-35.
https://doi.org/10.1016/0167-2789(84)90245-8.
9. Hunter C, Wei X. Coronovirus sequence assembly isolated from 2019/2020
Wuhan Outbreak Patient. Avalaible from: https://www.ncbi.nlm.nih.gov/
nuccore/LR757996?fbclid=IwAR1LXMPn5NSi6iNpz72cQYA2JwsK4gEvlI
MLp-yk_9KRTBG9sbVoezxiwA0. (last acess: 24.March 2020)
10. Available
from:
https://www.laconicml.com/predict-infected-people-coronavirus-python/?fbclid=IwAR2mbc_0jRFJRsuotKzfk_
T4WSfYZl7mR5wG34ufeIiGI-OrZHonkJ65s1k. (last acess: 24. March 2020)
REFERENCES
11. Available from: https://www.github.com/datasets/covid-19/commit/3e531107ded2ffeed441d39804b2e82cd14d88e7. (last acess: 24 March 2020)
1. Yang Z, Zeng Z, Wang K, Wong SS, Liang W, Zanin M, et al. Modified SEIR
and AI prediction of the epidemics trend of COVID-19 in China under public
health interventions. J Thorac Dis 2020;12(2):165-74.
12. Available from: https://www.kaggle.com/imdevskp/corona-virus-report.
(last acess 24 March 2020)
2. Hu Z, Ge Q, Li S, Jin L, Xiong M. Artificial Intelligence Forecasting
of COVID-19 in China. 2020. Available from: http//:www.arXiv.
Org>q-bio>arXiv:2002.07112.
13. Available from: https://www.worldometers.info/coronavirus/coronavirus-cases. (last acess 24 March 2020)
38