International Journal of Performance Analysis in Sport, 2021, 17(3): 299-306 doi: 10.23940/ijpe.21.03.p5.299306

Original article

LSTM and RNN to Predict COVID Cases: Lethality’s and Tests in GCC Nations and India

Razia Sulthana A.,a,*, Arokiaraj Jovithb, Jaithunbi A. K.c

Department of Computer Science, Birla Institute of Technology and Science Pilani (Dubai Campus), Dubai, 345055, UAE

Department of Information Technology, SRM Institute of Science and Technology, Kattankulathur, 603203, India

Department of Computer Science and Engineering, RMD Engineering College, Kavaraipettai, 601206, India

*Corresponding Author(s): Corresponding author. E-mail address: Corresponding author. E-mail address:


The spread of COVID across world countries is better handled by applying learning algorithms. Machine learning and deep learning algorithms can be applied to analyze the effects of COVID in multidimensional ways. This paper brings a detailed study of the COVID cases, deaths and tests across five of the GCC countries and India. The proposed method analyzes the COVID count against the population density of each of the countries. An analysis of the raw count would only give a false impression, whereas a population-based comparison gives the exact measure of the effect of COVID. As India is a densely populated country, the number of precautionary steps taken by the country against the population count needs to be measured for accurate prediction. Recurrent Neural Network and Long Short-term memory are used to predict the future cases, deaths and tests of India. A time span of 20 days is used in the prediction. In the sense that ith day to (i+20)th day values are taken to predict the (i+21)thday values. The accuracy of the LSTM model designed with multiple hidden layers is notable and the prediction error is minimal.

Keywords: long short-term memory ; recurrent neural network ; COVID ; Gulf corporation ; India

PDF (410KB) Metadata Related articles Export EndNote| Ris| Bibtex

Cite this article

Razia Sulthana A., Arokiaraj Jovith, Jaithunbi A. K.. LSTM and RNN to Predict COVID Cases: Lethality’s and Tests in GCC Nations and India. International Journal of Performance Analysis in Sport, 2021, 17(3): 299-306 doi:10.23940/ijpe.21.03.p5.299306

© 2021 Totem Publisher, Inc. All rights reserved.

1. Introduction

A virus [1] is a parasite that can multiply itself until it comes in contact with a susceptible cell. A virus contains nucleic acid in its centroid and a protein shell covering the centroid. The virus has either RNA (ribonucleic acid) or DNA (deoxyribonucleic acid) as genetic material [2]. Both these viruses replicate but in different intensities. In general, the rate of mutation in RNA virus is much higher than the rate of mutation in DNA virus and that makes RNA virus more resistant to drugs and so creates a challenge for scientists to create drugs or vaccine against it. Examples of DNA viruses include herpesvirus, parvovirus and RNA virus includes HIV virus and Coronavirus.

Coronavirus is an RNA viruses that cause respiratory tract infections. They cause severe health issues in humans and mammals. Their level of infection varies across different mammals. The coronavirus belongs to a family of Coronaviridae and can be classified into four genera [3]. The different varieties of coronavirus [4] are Severe acute respiratory syndrome (SARS), Middle East respiratory syndrome (MERS) also called as Camel flu and Coronavirus disease 2019 (COVID-19). The outer layer of the coronavirus has a spike protein that is the most important protein membrane of coronavirus. This spike protein, S protein, plays a vital role in the virus’ entry into a target cell. From January 2020 to now, the virus created a huge impact on the economy of the entire world population. The biotechnology world has developed a vaccine to inactivate this novel coronavirus.

A number of machine learning and deep learning algorithms have been proposed so far to predict the growth rate and death rate of COVID-19. The literature section discusses the existing works related to the prediction of COVID-19 from different perspectives using learning algorithms. The motivation behind choosing the work is to apply LSTM and RNN in analyzing the state of COVID to reduce the fatality rate. The proposed work analyzes the effect of COVID-19 in Gulf countries and in India. The contributions of this work include:

1) To predict the COVID growth rate in UAE, Bahrain, Kuwait, Qatar, Saudi Arabia and India.

2) To predict the COVID death rate in UAE, Bahrain, Kuwait, Qatar, Saudi Arabia and India.

3) Analyzing the rate of COVID tests taken with respect to population in each of the countries: UAE, Bahrain, Kuwait, Qatar, Saudi Arabia and India.

4) Predicting the number of Total Covid Cases, Covid Deaths and Covid Tests Using RNN and LSTM.

5) A multidimensional analysis is made to identify the countries that were stable during COVID

The article is organized as follows. Section 2 details the literature study; section 3 discusses the materials and methods applied in the proposed system; section 4 shows the prediction results using RNN and LSTM; section 5 concludes the study.

2. Literature Study

2.1. World Economy and COVID

GDP is the monetary value of the total goods produced by the country during a specific period. It’s just a scorecard of a country’s economic status among world countries. It is calculated on a quarterly and annual basis. Every fiscal year, a country’s government would release its financial reporting and budgeting. Economists decide over a country’s growth or recession using the GDP. GDP can be measured using a number of ways including Nominal, Real, Growth Rate, Per Capita. Of all these, the measure that mostly influences GDP is per capita. GDP per capita is the GDP of a single person in that country. The major factors that decide a countries stability would be gross domestic product (GDP) and poverty [5].

The fluctuations in the health care expenses also affect the GDP [6] of a country. However, a pandemic disease would badly disturb the country’s GDP. Clear speculation about the future and the early remedial steps taken to control such pandemics in the initial stage would help in saving the country’s economy from falling. At times, apocalyptic predictions may hit the country’s GDP badly leading to sudden inflation.

The FDI has a positive relationship with GDP per capita [7] of the United Arab Emirates (UAE). This article ascertains that inflation will increase the FDI within a certain limit. Yet, beyond a certain level inflation would disturb the GDP per capita and lead to extreme poverty.

2.2. Arab Countries and COVID

The maximum number of COVID cases among the Arab countries is seen in Kingdom of Saudi Arabia (KSA) [8]. Also, in the above article, a detailed analysis of STEMI (ST-Elevation Myocardial Infarction) and COVID is analyzed with the information obtained from 16 centers of Kingdom of Saudi Arabia from January 01 to April 30, 2019. The article [9] gives a detailed impact of COVID outburst in KSA. The author in [10] has forewarned that the Indoor Air Quality (IAQ) might lead to other health issues because of home isolation.

A detailed estimation of COVID cases in Iran based on the air travel overseas cases is shown in [11]. Binomial distribution and likelihood estimation is applied to predict the future cases using a 95% confidence level and by varying the size of the detection window. Across the world, mass gatherings were canceled as a de facto rule to prevent the spread of COVID [12]. A rapid review article [13] discusses the initial days of COVID spread and the fatality rate in Eastern Mediterranean region. One other article [14] reviews the awareness and attitude of the people during the COVID days and the practices followed to combat the spread of it.

2.3. Machine Learning and Deep Learning Methodologies Applied over COVID Prediction

A state-of-art of applying artificial intelligence and machine learning techniques for COVID is detailed in [15]. Prediction and forecasting of COVID, the vaccination, and drug towards curing COVID are also detailed. In another article [16], analysis of X-Ray images of COVID patients using Convolution Neural Networks is proposed. An Auxiliary Classifier Generative Adversarial Network (ACGAN) is modeled to create imitated X-Ray images that resemble those of COVID infected patients. Generative Adversarial Network (GAN) synthetic images can be generated to study the pattern of a disease that has minimal test data. The actual COVID images and generated COVID images are trained and tested separately and the system performance is measured. The generated images show better prediction and help us anticipate the X-Rays of COVID infected patients.

COVID control strategies such as applying population compartmentalization, extending the lockdown period, and raising the immunity level against the number of COVID cases are studied in [17]. It proposes a deep learning model using ANN using incremental learning technique for optimal decision making. The model can rebuild by adapting itself to the changing external factors that cause the COVID toll to rise. Another article proposes to drive through pharmacies to be deployed to avail medicinal facilities to all the people with zero waiting time [18]. This system might reduce the spread of COVID as most people visit pharmacies during the pandemic.

2.4. COVID’s Impact over Economy

The impact of COVID is likely to reduce the GDP of Indian economy by 10-31% [19]. The author has also discussed that the power sector would further weaken as there would be a severe reduction in the demand for power. On the other hand, the CO2emission is also expected to reduce. The aforementioned research proposed a system of linear input-output equations to predict the deviations from the actual behavior. In the same direction, another research article [20] has discussed the reduction in the demand for power and a drop in the concentration of harmful atmospheric gases during the COVID lockdown period. Article [21], analyzes the meteorological impact of COVID in India. A number of parameters like temperature, humidity, Aerosol Optical Depth (AOD) and NO2 are compared with a varying period of lockdown.

The impact of age and demographic information of the people is evaluated against the spread of COVID in [22]. The different sources of infection are studied and the potential factors that affect the number of infections is also studied. A multiple linear regression model followed by clustering is deployed to predict the future number of infections and the fatality rate.

Article [23] analyzes the number of COVID cases against the lockdown initiated by the government. Descriptive statistical parameters were evaluated to predict the COVID cases during the lockdown. Normal correlation and Tukey Lambda correlation were applied and the model was evaluated using sum of squares, t-ratio, and F-ratio.

This proposal analyzes Gulf Cooperation Council (GCC) countries and India in the following perspectives: Analysis of COVID in multiple perspectives, COVID growth rate, COVID death rate, and COVID tests taken across these GCC nations and India except Oman. We apply LSTM and RNN to predict the number of COVID deaths and COVID infected people. An analysis of the number of COVID tests taken to the respective country’s population is made. However, the countries that were stable during COVID are also identified.

3. Materials and Methods

3.1. Data

The data is taken from ( The COVID data of Arab countries namely, Kingdom of Saudi Arabia (KSA), United Arab Emirates (UAE), Bahrain, Kuwait, Qatar and India. The above Middle east countries’ population density, GDP per capita, cardiovascular death rate, diabetes prevalence, hospital beds and life expectancy were all analyzed before working with COVID chart. Population Density and GDP per capita of these countries show the eminence of these countries (Figure 1). Though the population density of Bahrain is more than India, it happens to maintain a worthy GDP, raising the economic level. However, India has eminent professionals in curing and controlling the cardiovascular death rate despite the low GDP. One notable feature is that, among all the countries assessed above, India is the only country whose extreme poverty status is 21.2, whereas for other countries it is 0.

Figure 1.

Figure 1.   Population Density, GDP per capita and cardiovascular death rate

However, it is very important to predict the number of beds in the country for accommodating the patients during emergency pandemic situations and life expectancy (Figure 2). Understanding this would help us distinguish which country provides the highest health care to its people.

Figure 2.

Figure 2.   Hospital beds per thousand and Life expectancy

As shown in Figure 2, the percentage of hospital beds is 28% in KSA and 2% in India. The population count in India is much higher and that would be the only cause for this reduced number. However, the life expectancy in all the countries is almost the same. It is justified and evidenced that the health care and doctors in India upheld and thus raising the life expectancy in India.

3.2. COVID Cases Growth Rate

The total number of COVID cases of these countries (Table 1) was assessed based on the day of occurrence from the first COVID case encountered in the respective countries (Figure 3). In all the countries, a steady increase in the COVID cases was found, so we narrowed and analyzed the growth rate from day 1 to day 180 with an interval of 20 days. 20 days of the interval is considered to be around 3 weeks which is a nominal period of break. It was observed that the growth rate of COVID in India and UAE heavily increased on the 60th and 80th days of COVID compared to other countries. However, the growth rate in other countries was not as bad as in India.

For the countries, after the 100th day, the COVID cases started to increase but not as bad as the initial days in those respective countries. During the initial 100 days, the increase on (i+1)th day was 15 times or 20 times the ith day. After the 100th day, the COVID cases of all the countries significantly increased, which is double times the previous day as seen in the chart. Yet, all the countries were able to reduce or stabilize the growth rates except India. Perhaps, the population in India would be a reason for that sake. The growth rate was brought to control in Bahrain, Kuwait, Qatar, and Saudi Arabia on the 160th or at least by the 180th day. Yet in India, the state is increasing exponentially and in UAE it is under control. Nevertheless, the comfort zone provided to people by UAE is far above that of other countries. For that reason, UAE has surpassed other countries in reducing COVID cases.

Table 1.   Total COVID cases across six countries

Ith DayUAEBahrainIndiaKuwaitQatarSaudi Arabia
Day 1111311
Day 2082113123501767
Day 40296434455629794934
Day 603332217107128921487227011
Day 804749581615712128604048170161
Day 1001473012311596622992175071119942
Day 12029485224071657994187997003197608
Day 140419903294136694656174107430258156
Day 160501414119074241768299did not reachdid not reach
Day 18057988did not reach1483156did not reachdid not reachdid not reach

New window| CSV

Figure 3.

Figure 3.   COVID death rate

Figure 4.  

3.3. COVID Cases Death Rate

After the outbreak of COVID in all these countries (Table 2), it was a mystery to predict the number of victims
(Figure 4). The number of deaths was minimal the first 60 days. After that, the death toll unexpectedly raised multiple times compared to previous days. Since the COVID flu takes 14 days to show its first symptom, it showed its effect by raising the death count by the 80th day. It remains a paradox over the number of COVID cases and deaths posted by the government as there may be many cases hidden without the public knowing. The death rate in all the countries was almost under control after the 100th day as the increase was (1/20)th of the previous day. Though India is very populated, it sounds good to know that it can control the death rate to a major level. For example, the number of fatalities on the 160th day was 20642 and on the 180th day was 33425, which states that India was able to curb the lethal cases.

Table 2   . Total COVID deaths across 6 countries

Ith DayUAEBahrainIndiaKuwaitQatarSaudi Arabia
Day 1000000
Day 20000001
Day 400401765
Day 6028291912184
Day 8028105079619379
Day 10013719198123669893
Day 1202456547063371151752
Day 140288109122373961602601
Day 16031814720642461did not reachdid not reach
Day 180342did not reach33425did not reachdid not reachdid not reach

New window| CSV

3.4. COVID Tests

The number of tests taken in these countries is analyzed to identify which country has taken more COVID tests on different days (Figure 5). Among all the countries, the number of COVID tests taken in India exceeds other countries; the next highest number of tests is in UAE. Nevertheless, the population density is less compared to others. UAE has taken trivial measures to control the spread of COVID by identifying those infected and isolating them from public. However, the population and COVID tests need to be mutually compared to unearth the truth behind which country has a higher number of tests taken.

Figure 5.

Figure 5.   shows a clear picture of the population in countries vs tests done. It clearly shows that UAE has taken a lot of tests compared to other countries and India has taken the least number of tests. India is slated to have the worst hit when compared to other countries.

Figure 6.  

4. Predicting the number of total COVID cases, COVID deaths and COVID tests using LSTM

Architecture of Artificial neural network (ANN) is Long Short-term memory (LSTM) [24] used in RNN [25]. Convolution Neural Network (CNN) [26-28] is another branch of machine learning used for processing images. It is used to predict the data in sequence. To predict the expected number of COVID cases in each country, instead of taking the interval as 20 days, almost all the days are taken into consideration. An example table (Table 3) is given below. It is true that COVID would heave outbreaks on different days in different countries [29].

Table 3.   Total COVID cases

Ith DayCountry
1st day of COVID .
2nd day of COVID .
3rd day of COVID .
4th day of COVID .
Nth day of COVID.

New window| CSV

Further, the RNN model using LSTM is trained in such a way that the predicted value is based on the previous 20 days data. With such style of prediction, daily training test is predicted and the accuracy of the model is tabulated below (Table 4).

Table 4.   LSTM error table

Adam Classifier: RNN using LSTM
Country: India
4 hidden layer, each with 10 units. A dropout of 0.2 in each layer.
Output layer with 1 unit
Calculated: mean_squared_error
epoch 1Total CasesTotal DeathsTotal tests
epoch 20.04630.05620.0334
epoch 30.03780.04980.0310
epoch 40.03610.04550.0298
epoch 50.03390.03440.0251
epoch 60.03250.03210.0201
epoch 70.02630.03140.0098
epoch 80.02240.02140.0077
epoch 90.01310.01210.0052
epoch 100.00570.00880.0001

New window| CSV

The prediction error was so minimal justifying that the proposed model sounds good for predicting sequential COVID data.

5. Conclusion

COVID is a pandemic disease and is highly contagious. The governments of the world have taken steps to curb the spread of the disease. A number of misconceptions arose among the people towards COVID and a number of initiatives are taken by different countries to save their people. The proposal works with 5 of the GCC nations and India. It analyzes the growth rate of total COVID cases, the death rate of COVID cases and the number of COVID tests taken by the country. The study compares the rate of impact with population density. All these factors are analyzed in detail with results tabulated. Another prediction analysis is made with RNN and LSTM to predict future cases, deaths and tests. This was implemented with a track day of 20 in count and the error was found to be too small. Hence RNN and LSTM together helps better predict sequential data.


Novel. The epidemiological characteristics of an outbreak of 2019

novel coronavirus diseases (COVID-19) in China

Zhonghua liu xing bing xue za zhi, 41(2), pp.145-151, 2020.

[Cited within: 1]


Prediction and design of DNA and RNA structures

New biotechnology, 27(3), pp.184-193, 2010.

[Cited within: 1]


Structure, function, and evolution of coronavirus spike proteins

Annual review of virology, 3, pp.237-261, 2016.

[Cited within: 1]

LiuJ., XieW., WangY., XiongY., ChenS., HanJ. and WuQ.

A comparative overview of COVID-19, MERS and SARS

International Journal of Surgery, 81, pp.1-8, 2020.

[Cited within: 1]

HaddadL.J. Nutrition and poverty, No. 576-2016-39115,2002.

[Cited within: 1]


The gross domestic product and health care spending

The New England Journal of Medicine, 369(2), pp.107-109, 2013.

[Cited within: 1]

AlshamsiK.H.and AzamM.

The impact of inflation and GDP per capita on foreign direct investment: the case of United Arab Emirates

Investment Management and Financial Innovations, 12(3), pp.132-141, 2015.

[Cited within: 1]

DaoulahA., HersiA.S., Al-FaifiS.M., AlasmariA., AljoharA., BalghithM., AlshehriM., YoussefA.A., ElSayedO., AlamaM.N.and RefaatW.A.

Stemi and COVID-19 pandemic in Saudi Arabia

Current problems in cardiology, p. 100656, 2020.

[Cited within: 1]

AmeenF., AmnaT., AlghamdiA.A., AlKahtaniM.D.and AlYahyaS.A.

Covid-19 pandemic outburst in Saudi Arabia: A glimpse

Saudi journal of biological sciences, 27(12), p.3547, 2020.

[Cited within: 1]


Indoor air quality and coronavirus disease (COVID-19

). Public Health, 2020.

[Cited within: 1]

ZhuangZ., ZhaoS., LinQ., CaoP., LouY., YangL. and HeD.

Preliminary estimation of the novel coronavirus disease (COVID-19) cases in Iran: A modelling analysis based on overseas cases and air travel data

International Journal of Infectious Diseases, 94, pp.29-31, 2020.

[Cited within: 1]

AhmedQ.A.and MemishZ.A.

The cancellation of mass gatherings (MGs)? Decision making in the time of COVID-19

Travel medicine and infectious disease, 34, p. 101631, 2020.

[Cited within: 1]

AlahM.A., AbdeenS. and KehyayanV. The first few cases and fatalities of Corona Virus Disease 2019 (

COVID-19) in the Eastern Mediterranean Region of the World Health Organization: A rapid review

Journal of Infection and Public Health, 2020.

[Cited within: 1]

AlahdalH., BasingabF. and AlotaibiR.

An analytical study on the awareness, attitude and practice during the COVID-19 pandemic in Riyadh, Saudi Arabia

Journal of infection and public health, 13(10), pp. 1446-1452, 2020.

[Cited within: 1]

LalmuanawmaS., HussainJ. and ChhakchhuakL.

Applications of machine learning and artificial intelligence for Covid-19 (SARS-CoV-2) pandemic: A review

Chaos, Solitons & Fractals, p. 110059, 2020.

[Cited within: 1]

WaheedA., GoyalM., GuptaD., KhannaA., Al-TurjmanF. and PinheiroP.R.

Covidgan: data augmentation using auxiliary classifier gan for improved covid-19 detection

Ieee Access, 8, pp.91916-91923, 2020.

[Cited within: 1]

FarooqJ. and BazazM.A.

A novel adaptive deep learning model of Covid-19 with focus on mortality reduction strategies

Chaos, Solitons & Fractals, 138, p.110148, 2020.

[Cited within: 1]

HussainR. and DawoudD.M. pandemic pandemic. Drive-thru pharmacy services: A way forward to combat COVID-19 Research in Social and Administrative Pharmacy, 17(1), pp. 1920-1924, 2021.

[Cited within: 1]


The COVID-19 lockdown in India: Impacts on the economy and the power sector

Global Transitions, 2, pp.150-156, 2020.

[Cited within: 1]

GhoshA., NundyS. and MallickT.K. How India is dealing with COVID-19 pandemic. Sensors International, 1, p.100021, 2020.

[Cited within: 1]


Effect of meteorological parameters on spread of COVID-19 in India and air quality during lockdown

Science of the Total Environment, 745, p.141021, 2020.

[Cited within: 1]


Potential of age distribution profiles for the prediction of COVID-19 infection origin in a patient group

Informatics in Medicine Unlocked, 20, p.100364, 2020.

[Cited within: 1]


Is the lockdown important to prevent the COVID-19 pandemic? Effects on psychology, environment and economy-perspective

Annals of medicine and surgery, 56, pp.38-42, 2020.

[Cited within: 1]

HochreiterS. and SchmidhuberJ.

Long short-term memory

Neural computation, 9(8), pp. 1735-1780, 1997.

[Cited within: 1]

MikolovT., KombrinkS., BurgetL., ČernockýJ. and KhudanpurS. Extensions of recurrent neural network language model. In 2011 IEEE international conference on acoustics, speech and signal processing ( ICASSP), pp.5528-5531, 2011.

[Cited within: 1]

SulthanaA.R., GuptaM., SubramanianS. and MirzaS.

Improvising the performance of image-based recommendation system using convolution neural networks and deep learning

Soft Computing, 24(19), pp.14531-14544, 2020.


V., Le, H.D. and Nhan, N.T. Vietnamese handwritten character recognition using convolutional neural network

IAES International Journal of Artificial Intelligence, 9(2), p.276, 2020.


Convolutional neural networks for leaf image-based plant disease classification

IAES International Journal of Artificial Intelligence, 8(4), p.328, 2019.

AroraP., KumarH. and PanigrahiB.K.

Prediction and analysis of COVID-19 positive cases using deep learning models: A descriptive case study of India

Chaos, Solitons & Fractals, 139, p.110017, 2020.

[Cited within: 1]