The Tank

This statement under collusion pretty much says everything about the computer models and the predictions.

"COVID-19 is still an unclear infectious disease, which means we can only obtain an accurate SEIR
prediction after the outbreak ends."

Now if we just had adults around the planet who are adult enough to admit they messed up we can roll back some of the B.S. legislation and get on with our lives. Preferably before the global economy collapses.

copied from:

https://www.who.int/bulletin/online_first/20-255695.pdf

DISCLAIMER This paper was submitted to the Bulletin of the World Health Organization and was posted to the COVID-19 open site, according to the protocol for public health emergencies for international concern as described in Vasee Moorthy et al. (http://dx.doi.org/10.2471/BLT.20.251561).

The information herein is available for unrestricted use, distribution and reproduction in any medium, provided that the original work is properly cited as indicated by the Creative Commons Attribution 3.0 Intergovernmental Organizations licence (CC BY IGO 3.0).

RECOMMENDED CITATION Binti Hamzah FA, Lau C, Nazri H, Ligot DV, Lee G, Tan CL, et al. CoronaTracker: Worldwide COVID-19 Outbreak Data Analysis and Prediction. [Submitted]. Bull World Health Organ. E-pub: 19 March 2020. doi: http://dx.doi.org/10.2471/BLT.20.255695

CoronaTracker: World-wide COVID-19 Outbreak Data Analysis and Prediction CoronaTracker Community Research Group

Fairoza Amira Binti Hamzaha, Cher Han Laub, Hafeez Nazric, Dominic Vincent Ligotd, Guanhua Leee, Cheng Liang Tanf, Mohammad Khursani Bin Mohd Shaibg, Ummi Hasanah Binti Zaidonh, Adina Binti Abdullahi, Ming Hong Chungj, Chin Hwee Ongk, Pei Ying Chewl and Roland Emmanuel Salungam

a The Kyoto College of Graduate Studies for Informatics b Lead c Media Prima Digital Sdn Bhd d Cirrolytix e National University Health System f International Medical University g Vase.ai h Institut Wanita Berdaya Selangor i University of Malaya j Quest International University Perak k ST Engineering l Universiti Malaysia Sarawak m National Capital Region, Philippines

Correspondence to Fairoza Amira Binti Hamzah (email: fairozaamira@gmail.com) (Submitted: 18 March 2020 – Published online: 19 March 2020)

CoronaTracker: World-wide COVID-19 Outbreak Data
Analysis and Prediction
CoronaTracker Community Research Group

Abstract
Background
COVID-19 outbreak was first reported in Wuhan, China and has spread to more than 50 countries.
WHO declared COVID-19 as a Public Health Emergency of International Concern (PHEIC) on 30
January 2020. Naturally, a rising infectious disease involves fast spreading, endangering the health
of large numbers of people, and thus requires immediate actions to prevent the disease at the
community level. Therefore, CoronaTracker was born as the online platform that provides latest
and reliable news development, as well as statistics and analysis on COVID-19. This paper is done
by the research team in the CoronaTracker community and aims to predict and forecast COVID
19 cases, deaths, and recoveries through predictive modelling. The model helps to interpret patterns
of public sentiment on disseminating related health information, and assess political and economic
influence of the spread of the virus.

Methods:

Real-time data query is done and visualized in our website, then the queried data is used for
Susceptible-Exposed-Infectious-Recovered (SEIR) predictive modelling. We utilize SEIR
modelling to forecast COVID-19 outbreak within and outside of China based on daily observations.
We also analyze the queried news, and classify the news into negative and positive sentiments, to
understand the influence of the news to people’s behavior both politically and economically.

Findings:

At the time of writing this paper, the number of confirmed cases is expected to exceed 76000 cases,
and reach the peak of this outbreak before 20 February 2020. The average Infected-Suspected ratio
was found to be 2.399 which we used to initialize the number of Exposed individuals as a product
of the number of Infected individuals on 20 Jan 2020. This outbreak is assumed to reach its peak
in late May 2020 and will start to drop around early July 2020. Based on the news queried in our
system, we found that there are more negative articles than positive articles, and displayed similar
words for both negative and positive sentiments. The top five positive articles are about
collaboration and strength of individuals in facing this epidemic, and the top five negative articles
are related to uncertainty and poor outcome of the disease such as deaths.

Conclusions:

COVID-19 is still an unclear infectious disease, which means we can only obtain an accurate SEIR
prediction after the outbreak ends. The outbreak spreads are largely influenced by each country’s
policy and social responsibility. As data transparency is crucial inside the government, it is also
our responsibility not to spread unverified news and to remain calm in this situation. The
CoronaTracker project has shown the importance of information dissemination that can help in
improving response time, and help planning in advance to help reduce risk. Further studies need to
be done to help contain the outbreak as soon as possible.

Keywords: COVID-19, data analysis, sentiment analysis, predictive modelling, SEIR

1. Introduction

On 31 December 2019, the first reported case in the COVID-19 outbreak was reported in
Wuhan, China. The first case outside of China was reported in Thailand on 13 January 2020 [1].
Since then, this ongoing outbreak has now spread to more than 50 other countries [2]. WHO
declares COVID-19 outbreak as a Public Health Emergency of International Concern (PHEIC) by
WHO on 30 January 2020 [3]. There are over 76,000 cases of confirmed COVID-19 worldwide as
of 20 February [2].

An infectious disease outbreak is the occurrence of a disease that is not usually expected
in a particular community, geographical region, or time period [4]. Typically, a rising infectious
disease involves fast spreading, endangering the health of large numbers of people, and thus
requires immediate action to prevent the disease at the community level [5]. COVID-19 is caused
by a new type of coronavirus which was previously named 2019-nCoV by the World Health
Organization (WHO). It is the seventh member of the coronavirus family, together with MERS
nCoV and SARS-nCoV, that can spread to humans [1]. The symptoms of the infection include
fever, cough, shortness of breath, and diarrhea. In more severe cases, COVID-19 can cause
pneumonia and even death [6]. The incubation period of COVID-19 can last for 2 weeks or longer
[7]. During the period of latent infection, the disease may still be infectious. The virus can spread
from person to person through respiratory droplets and close contact [8].

An ‘infodemic’ has accompanied the COVID-19 outbreak which is essentially an over
abundance of information regarding the outbreak. As some of the information available to the
public may not be accurate, it becomes hard for people to find reliable sources and trustworthy
guidance when they need it [9]. Because of the high demand for appropriate and trustworthy
information about 2019-nCoV, WHO technical risk communication and social media teams have
been working closely to track and respond to myths and rumors via its headquarters in Geneva, its
six regional offices and its partners. The organization is working continuously to identify the most
widespread rumors that can possibly harm the public’s health, such as inaccurate prevention
measures or claims of cures. These myths are then rebutted with evidence-based information. WHO
is making public health information and advice on the COVID-19, including myth busters,
accessible on its social media channels (including Weibo, Twitter, Facebook, Instagram, LinkedIn,
Pinterest) and on their website [10].

Communication during emerging pandemics presents a distinctive public health
education task. Health consumers must be informed about an impending health threat [11].
However, there may be difficulties in providing accurate information regarding the outbreak in the
initial stage. This is mainly related to the high degree of uncertainty about the exact route of
transmission, treatment of the infections, and prospects of recovery in an outbreak. All countries
need to prepare existing public health communication networks, media and community
engagement staff for a possible case in their country, as well as for the appropriate response if it
happens. The governments should coordinate communications with other response organizations
and include the community in response operations. WHO stands ready to coordinate with partners
to support countries in their communication and response to community engagement.
To ensure a people-centered response to COVID-19, an expanding group of global
response organizations such as the United Nations Children’s Fund (UNICEF) and the International
Federation of Red Cross and Red Crescent Societies (IFRC) are coordinating efforts with WHO to
apply biomedical recommendations at the community level. These organizations are active at the
global, regional and country level to ensure that affected populations have a voice and are part of
the response. Ensuring that global recommendations and communication are tested and adapted to
local contexts will help countries to gain better control over the COVID-19 outbreak [10].

Peoples’ response to the news about a spreading contagious disease is likely to lead to
increased anxiety and amplification of risk perceptions [11]. Social media networks have
functioned as channels for firsthand information from which the public can acquire disease-related
information during infectious disease outbreaks. These platforms also enable simple and quick
sharing of information with family, friends, and neighbors in real time [12]. For example, the
Ministry of Health in Malaysia have been uploading posts related to COVID-19 to educate the
public since 19 January [13] and their Director General of Health is also active on his own
Facebook page to clear confusion and doubts for the public [14]. This is important when traditional
forms of media are unable to provide relevant and timely information to the public. Social media
now serves as a major, immediate information source but while the focus of latest information has
been on the role of social media during infectious disease outbreaks, the question that should be
brought to light is still, how the use of social media may trigger the public’s emotional or
noncognitive response, affect perception of risk, and preventive behaviors [10].

Therefore, CoronaTracker [15] was born as the online platform that provides the latest
and verified news development, statistics and analysis on COVID-19. This platform is a
community-based project initiated on 25th February out of concerns on the outbreak that halted
Mainland Chinese of Lunar New Year’s celebration. The CoronaTracker website was launched on
27th January 2020, after two days working relentlessly, and has gathered more than 1300 volunteers
across the globe. This paper is a part of a work by the research team of CoronaTracker community.
The main objective of this paper is to predict and forecast COVID-19 cases, deaths, and recoveries
through predictive modelling, and to decipher patterns on public sentiment related to health
information dissemination. At the same time, assess the political and economic impact of the virus
spread.

We propose a comprehensive framework to manage health information data as a tool for
public health practitioners in managing epidemics and crafting public health response and policy.
This study focuses on the role of audiences in the process of disseminating health risk information
and examines behaviors that contribute to information amplification upon hearing the news.
The structure of this paper is as follows; Section 1 introduces COVID-19 and
CoronaTracker community, as well as explains the significance of this research. Section 2 describes
on related works in predictive modelling of the paper and news-based sentiment analysis for this
research on psychological, politics and economics aspects. Section 3 explains our study design and
methodologies. Section 4 presents our findings in current trends, predictive modelling and
sentiment analysis of the outbreak. Our findings are discussed in Section 5 and this paper is
concluded in Section 6.

2. Related Works

SEIR refers to Susceptible, Exposed, Infectious, and Removed or Recovered, respectively.
It is based on the SIR model but adds the Exposed compartment as a variable. Susceptible refers
to individuals who can catch the infection and may become hosts if exposed, Exposed are
individuals who are already infected but are asymptomatic, Infectious are individuals who are
showing signs of infection and can transmit the virus, Removed or Recovered are individuals who
are previously infected but are no longer infectious and already immune to the virus [16].
Once the compartments of SIR or SEIR models are determined, modelling can be done
using a variety of methods. In [17], a Conditional Autoregressive (CAR) was used to account for
epidemics with a spatial or transportation-related vector and modelled with MCMC. In [16],
demographic effects such as birth and death rates were added to the SEIR to model equilibria with
vital dynamics.

Sentiment analysis is a supervised machine learning problem. There are different types
of sentiment analysis including fine-grained sentiment analysis, emotion detection, aspect-based
sentiment analysis and multilingual sentiment analysis. In binary sentiment classification, the
possible categories are positive and negative. In fine-grained sentiment classification, there are five
groups (very negative, negative, neutral, positive, and very positive). Sentiment analysis is one of
the most popular tasks in natural language processing, and there has been a lot of research and
progress in solving this task accurately [18].

Deep neural networks are widely used in sentiment polarity classification; however, it
often requires huge numbers of training data, and the size of training data varies quite significantly
among domains. In [19], it was found that a dual-module approach is the best method that
encourages the learning of models with promising generalization abilities. Bidirectional Encoder
Representations from Transformers (BERT) is an embedding layer designed to train deep
bidirectional representations from unlabeled texts by jointly conditioning on both left and right
context in all layers. It is pretrained from a large unsupervised text corpus such as Wikipedia or
BookCorpus. There are 15% of the words in the input sequence are masked out which is one of the
objectives of BERT. Then, a deep bidirectional Transformer encoder is fed by the entire sequences
so that the model learns to predict the masked words.

Moreover, this small model has been trained on SST-2 dataset which is a common dataset
for sentiment-analysis [20]. However, there are few disadvantages in this method as it is based on
SST-2 dataset which is for movie reviews and our dataset is about coronavirus news. It is a similar
task which is for sentiment analysis but it does not perform that well because sentiment for movies
and news might be different. However, it is the fastest way to get results and act as a benchmark or
starter for further research. It can also be easily improved by adding more dataset for our domain
(coronavirus news). Last but not least, it can do prediction instantly compared to previous methods
that need bigger computer resources.

3. Methods
3.1 Data Source

Data is extracted from verified sources such as John Hopkins University [21], WHO and
DingXiangYuan, a website authorized by the Chinese government. The sites reported confirmed
COVID-19 cases, as well as recovered and deaths for affected countries and regions. Details on
how our team fetched the data is in Section 3.2.

3.2 Data Visualization

The data collected in CoronaTracker is available on data lakes platform. Both the
platform and dashboard are hosted in Amazon Web Services (AWS). We provisioned AWS
Relational Database Service (RDS) to host the data in MySQL table form. All the data collected
and ingested using Python program running in AWS Elastic Compute Cloud (EC2) and was
scheduled to automatically update every 15 minutes. The size of the database is relatively around
30GB.

Fig. 1 High level diagram ingestion for CoronaTracker
We developed our own micro-services and scraper in Python to fetch the data and news
from the sources. Python has been the best open source platform to use because of the less level of
complexity, readily libraries and easy to deploy in production environment (EC2). The code snippet
for our data fetcher is shown as in Fig. 2.

Fig. 2 Code snippet for data fetcher

The fetched data will be stored in relational database, MySQL. The micro-services will
crawl the data every 15 minutes and store it in table form. The data mostly will be stored in raw
format before being aggregated for visualization. Aggregate functions are built using SQL
statement like below snippet. Aggregation is important to show latest of the sum values for every
country in the table.

Fig. 3 SQL statement and AGG (aggregate) functions
Example of the data in the databases is shown in Fig. 4. We also geo-coded the location
for easier mapping during visualization.


Fig. 4 Result from AGG query

For every 15 min, the cumulative case counts will be updated for all provinces and other
affected countries. In the beginning, we found that WHO and JHU data are relatively slow
compared to other sources, thus we implement manual update to our system after verifications to
allow more real-time data.

3.3 Predictive Modelling – SEIR Model

Here we will briefly discuss the properties of basic Susceptible-Exposed-Infected
Removed (SEIR) system that will be used to describe the recent outbreak of COVID-19 in China
[22]. We considered a simple SEIR epidemic model for the simulation of the infectious-disease
spread. Individuals were each assigned to one of the following disease states; Susceptible (S),
Exposed (E), Infections (I) and Removed (R) which refers to segment not yet infected and disease
free, individuals that are experiencing incubation duration, the confirmed (isolated) cases,
recovered individuals, respectively. The SEIR diagram in Fig. 5 shows how individuals move
through each compartment in the model.

Fig. 5 SEIR model with four states [23]

Parameters within this model are:
1. β, controls the rate of spread, which represents the probability of transmitting disease between
a susceptible and an infectious individual [24]. The reproductive number used in this paper is

2.2.
2. Incubation rate σ, is the rate of latent individuals becoming infectious. Given the known
average duration of incubation Y, σ = 1/Y [24]. The average incubation duration of 5.2 days
are used here.

3. Recovery rate γ = 1/D, is determined by the average duration of recovery D, of infection. After
this period, they enter the removed phase. The average duration of infection is calculated as
the average serial interval minus the average incubation duration. The average serial interval
of 7.5 days is used in this paper [24]. 2.3 days of an average infectious duration is used here.
Fig. 6 shows the diagrammatic representation of virus progress in an individual, where
infectious occurs at 𝑡, during the latent period, infected individual is not infectious, and at 𝑡,
symptoms appear [16]. The first transmission to the left healthy individuals is at 𝑡. After 𝑡, the
removed (recovered) people are considered no longer infectious.

Fig. 6 Virus progress in an individual by using the SEIR model
We can describe the virus transmission by the following nonlinear ordinary differential
equation [23] as shown in equation (10) to (13).

= −

, (10)

=

− 𝜎𝐸, (11)

= 𝜎𝐸 − 𝛾𝐼, (12)

= 𝛾𝐼, (13)

3.4 Sentiment Analysis

After we have stored the news inside the CoronaTracker database, we extract news description
as it contains a summary of the news that is neither too short nor too long, which can be bad for
the model we are going to use otherwise. We only select descriptions that are at least more than 8
words, and discard non-English descriptions because the pre-trained model we use have been
trained on SST-2 [25], which is a dataset for sentiment analysis for English language. We use a
library called transformers by huggingface [26]. The input sentences will be separated by their
respective polarity for further analysis like topic modelling and generating word cloud for each
polarity.

4. Findings
4.1 Current Outbreak Trends
Fig. 7 to Fig. 9 shows the current trends for COVID-19 outbreak as displayed in CoronaTracker
website [27]. The cases reported are visualized in analytics dashboard to show the outbreak trend
for confirmed, recovered and deaths cases for all regions and countries. This aligns with our
objectives to show the outbreak progress over the period of time for each segment. It was found
that the total number of confirmed cases for all countries and regions are increasing steadily, but
on day 24, the huge increment with 15000 differences were discovered. This is due to the change
in how China measured the confirmed cases. Fig. 9 is plotted by using an open source mapping
library from Leaflet [28]. Table 1 shows the details on reported cases according to the countries
[27]. Both are retrieved on 3rd March 2020.

Fig. 7 Outbreak trend over time


Fig. 8 Most affected regions


Fig. 9 World map of affected region, where the darker the redness in the map, the more the
number of infected cases detected.
Table 1 Reported cases as of 3rd March 2020
Country Total Confirmed Total Recovered Total Deaths
China 80,151 47,367 2,945
South Korea 5,186 30 28
Italy 2,036 149 52
Iran 1,501 291 66
Others* 706 10 6
Japan 274 32 6
France 191 12 3
Germany 165 16 0
Spain 120 2 0
Singapore 108 78 0
United States 105 7 6
Hong Kong 100 37 2
Kuwait 56 0 0
Bahrain 49 0 0
Thailand 43 31 1
Taiwan 42 12 1
Switzerland 42 0 0
United Kingdom 40 8 0
Malaysia 36 22 0
Australia 31 11 1
Canada 27 6 0
Iraq 26 0 0
Norway 25 0 0
United Arab Emirates 21 5 0
Netherlands 18 0 0
Austria 18 0 0
Vietnam 16 16 0
Sweden 15 0 0
Lebanon 13 0 0
Israel 12 1 0
Macau 10 9 0
Iceland 9 0 0
Croatia 9 0 0
Belgium 8 1 0
San Marino 8 0 0
Greece 7 0 0
Qatar 7 0 0
Finland 6 1 0
Oman 6 1 0
Ecuador 6 0 0
India 5 3 0
Mexico 5 1 0
Algeria 5 0 0
Pakistan 5 0 0
Denmark 4 0 0
Czech Republic 4 0 0
Russia 3 2 0
Philippines 3 2 1
Georgia 3 0 0
Romania 3 0 0
Azerbaijan 3 0 0
Egypt 2 1 0
Brazil 2 0 0
Portugal 2 0 0
Indonesia 2 0 0
Cambodia 1 1 0
Nepal 1 1 0
Sri Lanka 1 1 0
Afghanistan 1 0 0
Estonia 1 0 0
Andorra 1 0 0
Nigeria 1 0 0
Luxembourg 1 0 0
Saudi Arabia 1 0 0
* Cases identified on a cruise ship currently in Japanese territorial waters.

4.2 Predictive Modelling
In this section, we will model the global trajectory of the infection counts using the SEIR
model, 240 days from the start date of 20 January 2020. This start date was chosen because earlier
dates were assumed as “burn-in” for the reporting of infection counts. The parameters of the SEIR
model were determined in section 3.2 and the world’s population are assumed to be 7.5 billion
people.
Using the data source from JHU, we lacked an initial number of Exposed individuals. To
gather this information, we made the estimation that the number of Infected individuals today is an
approximate ratio to the number of suspected individuals 6 days ago with an incubation duration
of 5.2 days; we rounded up the incubation duration from 5.2 days for this calculation. The data to
calculate this ratio was collected via real-time query from DingXiangYuan which provided
suspected and infected counts in mainland China. Using two weeks of data from 20 Jan 2020, the
average Infected-Suspected ratio was found to be 2.399 which we used to initialize the number of
Exposed individuals as a product of the number of Infected individuals on 20 Jan 2020.
Using Scipy’s implementation [29] of the numerical integration of ordinary differential
equations, odeint, we plotted the E and I trajectories of the SEIR compartment. The Exposed and
Infected trajectories tell us the number of individuals in those compartments over time. It can be
seen from the plot that the maximum number of Infected individuals is 425.066 million globally
on 23 May 2020. Thereafter, the number of Infected individuals dropped to under 10 million on 12
July 2020, under 1 million on 3 Aug 2020 and under 10,000 at the end of the trajectory on 14 Sep
2020.

Fig. 10 Simulated Exposed and Infected Per Day
Although the SEIR model is a numerical simulation, the numbers provide us a degree to
which the COVID-19 cases can surge to. These trajectories could serve as a means for governments,
businesses and individuals to plan and mitigate for such a spike in Infected cases. Everyone should
work towards blunting the curve and stopping the spread as per instructions set out by their
governments on personal hygiene, control measures and refraining from mass gatherings.
4.3 Sentiment Analysis
This paper also presents a sentiment analysis of recent verified news to understand
peoples’ reaction psychologically, politically and economically. Table 2 shows the number of
positive and negative articles analysed from the news.
Table 2 Number of positive and negative articles from the news
Sentiment n Sentiment n
Positive 561 Negative 2548
Fig. 11(a) and (b) show the word cloud for positive and negative sentiments, respectively.
It was found that both sentiments have the word “China” but it is necessary to check the word
sequences to identify whether sentences that contain the word “China” are considered positive or
negative.

(a) Positive sentiments

(b) Negative sentiments
Fig. 11 Word cloud for sentiments analysis on the news
Top 5 positive and negative articles are represented in Table 3 and 4.
Table 3 Top 5 positive articles
News No. of words Sentiment type Score
MELBOURNE: The coach of the Chinese
women's soccer team has praised the "strong
hearts" of his players after they emerged
undefeated from an ...
24 POSITIVE 0.999845
WHO Director-General said solid collaboration,
transparency, prompt sharing of data, and
accurate ad
13 POSITIVE 0.999836
Chinese President Xi Jinping has written a letter
expressing thanks to the Bill & Melinda Gates
Foundation for the organisation's "generosity"
and ...
23 POSITIVE 0.999809
SINGAPORE: The reaction of Singaporeans, as
they pitch in to help with the Wuhan coronavirus
threat, has been 'amazing', said Law and Home
Affairs Minister K Shanmugam on Sunday (Feb
2).
31 POSITIVE 0.999795
"We very much appreciate the efforts of the
doctors. They took good care of us," said a
patient surnamed Li while walking toward her
35 POSITIVE 0.999698
family after bidding farewell to the medical
workers in Hongshan Gymnasium.

Table 4 Top 5 Negative Sentiments Articles
News No. of words Sentiment type Score
SINGAPORE: The Wuhan coronavirus will
cause current economic uncertainties to
escalate, but Singapore's economy is prepared
to weather the impending financial impact, said
Manpower Minister Josephine Teo on
Thursday (Jan 30).
31 NEGATIVE 0.966091
BEIJING - A Chinese teenage boy suffering
from cerebral palsy was found dead at home in
Hubei province six days after his father and
younger brother were quarantined for
suspected infection with a novel coronavirus..
Read more at straitstimes.com.
39 NEGATIVE 0.936342
An image of a Chinese man collapsed in a
Seoul subway station, apparently too sick to
stand, has been held up by many Koreans as
proof of just how great a threat the
33 NEGATIVE 0.99643
It is not necessary to wear a face mask for now
since no human-to-human transmission of the
novel coronavirus (2019-nCoV) has occurred
in Malaysia, says infe...
26 NEGATIVE 0.771767
US researchers are already working on a
vaccine against the new virus that has emerged
in China.
17

5. Discussion
5.1 Health information dissemination
Information from news articles played an important role in empowering citizens during
an epidemic. However, our analysis found more negative articles than positive articles which is a
concern. (2548 vs. 561) The word clouds corresponding to the negative and positive statements
displayed similar words. A content analysis of text mined Ebola on news articles and scientific
publications showed similar findings. According to An’s study, they found a difference in coverage
of topics but a word co-occurrence map shared similar entities, which showed the limitation of
single word tagging for sentiment analysis during an epidemic. An interface for citizens to record
their sentiments and expressing their opinions on the news articles could contribute towards
improving these findings [30]. The top 5 positive articles were about collaboration and strength
of individuals in facing this epidemic, whereas the top 5 negative articles were related to
uncertainty and poor outcome of disease like death. Another study evaluating information shared
during the Ebola epidemic also found news articles to place more focus on event-related entities
such as person, organization and location, while social media information like in Twitter places
covers more time-oriented entities [31].
5.2 Economic and politics impact
COVID-19 is spreading with astounding speed and have severe consequences. The cases
increase rapidly with major outbreaks in South Korea, Italy, Iran, the United States and more than
50 other countries. Governments are put to the test with escalating high-pressure and costs of the
outbreak in terms of public trust and the economy. Any mismanagement could carry political costs,
as their legitimacy and competency will be called into question by the people.
Transparency is crucial and starts inside the government [32]. When the first outbreak
began, the public reacted to transparency, comprehensive and timely information and data provided.
Withholding the information created a vicious cycle of mistrust in authorities [33].
Clear direction in mitigating the outbreak is a must. The accurate and reliable information
are underpinned to ensure the containment effort is aggressively used and disseminated at any form
of social media to alert the public. In the absence of facts and trust, rumours and panic are inevitable
as people turned out to be emotional. These emotions created anger and fear that posed a threat to
the government. According to Amesh Adalja, a senior scholar at the Johns Hopkins Center for
Health Security, argued that the public is discouraged to view the response from the government
as a failure as that could create distrust of future public health measures [34].
At the geopolitical level, travel bans, border closing, trade controls, and others are other
containment measures to urgently enhance global readiness needed in response to Covid-19. The
consequences of outbreaks and epidemics are not distributed equally and the economic impact is
also highly uncertain. Some sectors may even benefit financially, while others will suffer
disproportionately [35]. Most countries that are badly affected by Covid-19 have prepared the
stimulus package or domestic economic growth plan in order to boost spending activities.
According to Wei Yao, the chief Asia economist at Société Générale has stated that the
worse the economic data are right now, the more aggressive policy responses will be. Given the
current development, the infrastructure stimulus is almost certain and the question is just how much
[36].

6. Conclusion
This research presented current trends of COVID-19 outbreak from 22nd Jan 2020 to 3rd
March 2020 as visualized in CoronaTracker website. The trajectory of the outbreak is also
forecasted by using SEIR model, 240 days from 20th January 2020. We also analyzed the sentiments
from news extracted by CoronaTracker to further understand people’s reaction towards this
outbreak. COVID-19 is still an infectious disease with some unclear or unknown properties, which
means accurate SEIR prediction can only be obtained once the outbreak has been successfully
contained. The outbreak spreads are largely influenced by each country’s policy and social
responsibility.
In a pandemic like this, providing timely information to the public is paramount. A
platform like CoronaTracker will assist the government and authorities to disseminate verified
articles, provide updates to the situation, and advocate good personal hygiene to the people.
CoronaTracker is built out of social responsibility to spread awareness to the common people by
providing scientific-based data analysis, prediction and verified news. Our website has attracted
more than 1000 users viewing our page at one time and surpassed 4 million-page views on 16th
March 2020. Our platform is not limited to CoronaTracker webpage, but also Facebook page,
Twitter and LinkedIn, with our ultimate goal is to provide verified news and current statistics to
the world.
This paper is still an ongoing research as many more investigations regarding this disease
can be carried out. Yet, it serves as the starting phase to research more in depth on questions that
revolve around this global pandemic.

Acknowledgments
This work is supported by Amazon AWS, LEAD and CirroLytix. We thank other
CoronaTracker core members, Tan Wei Seng, Poon Chee Him and Marcus Chia for managing the
community and the website. We also thank Yiran Jing, Zi Qing Ang, Goh Kok Han, Nikky Ng,
Debby Huang, Shi Yong Pang, Roland Emmanuel Salunga and Kenny Kang for their involvement
in our team discussion.

Author Group and Contributions
The CoronaTracker Research Team includes Fairoza Amira Binti Hamzah, Cher Han Lau,
Hafeez Nazri, Dominic Ligot, Ummi Hasanah Zaidon, Mohammad Khursani Bin Mohd Shaib,
Guanhua Lee, Tan Chen Liang, Adina Binti Abdullah, Chung Ming Hong, Chin Hwee Ong, and
Chew Pei Ying. All team members jointly conceptualized the study, analyzed and interpreted that
data, wrote and revised the manuscript and decided to submit for publication.
Corresponding author: Fairoza Amira Binti Hamzah,

References

[1] D. Hui et al, "The continuing 2019-nCoV epidemic threat of novel coronavirus to global
health - The latest 2019 novel coronavirus outbreak in Wuhan, China," International Journal
of Infectious Diseases, vol. 91, pp. 264-266, 2020.
[2] World Health Organization (WHO), "Coronavirus disease 2019 (COVID-19) Situation
Report - 35," WHO, 2020.
[3] World Health Organization (WHO), "Statement on the second meeting of the International
Health Regulations (2005) Emergency Committee regarding the outbreak of novel
coronavirus (2019-nCov)," WHO, 2020.
[4] D. Toppenberg-Pejcic, J. Noyes, T. Allen, N. Alexander, M. Vanderford and G. Gamhewage,
"Emergency Risk Communication: Lessons Learned from a Rapid Review of Recent Gray
Literature on Ebola, Zika, and Yellow Fever," Health Communication, vol. 34, no. 4, pp. 437
455, 2018.
[5] L. Lin, R. McCloud, C. Bigman and K. Viswanath, "Tuning in and catching on? Examining
the relationship between pandemic communication and awareness and knowledge of MERS
in the USA," Journal of Public Health, p. fdw028, 2016.
[6] WHO, "Advice for Public," WHO Int., 2020. [Online]. Available:
https://www.who.int/emergencies/disease ... for-public.
[Accessed 27 February 2020].
[7] World Health Organization, "Report of the WHO-China Joint Mission on Coronavirus
Disease 2019 (COVID-19)," World Health Organization, 2020.
[8] "Coronavirus Disease 2019 (COVID-19)," Centers for Disease Control and Prevention,
2020. [Online]. Available: https://www.cdc.gov/coronavirus/2019
ncov/about/symptoms.html. [Accessed 27 February 2020].
[9] World Health Organization, "2019 Novel Coronavirus (2019-nCoV): Strategic Preparedness
and Response Plan," World Health Organization, Geneva, 2020.
[10] World Health Organization, "Coronavirus disease 2019 (COVID-19) Situation Report- 13,"
World Health Organization, 2020.
[11] C. Chew and G. Eysenbach, "Pandemics in the Age of Twitter: Content Analysis of Tweets
during the 2009 H1N1 Outbreak," PLoS ONE, vol. 5, no. 11, p. e14118, 2010.
[12] S. Oh, S. Lee and C. Han, "The Effects of Social Media Use on Preventive Behaviors during
Infectious Disease Outbreaks: The Mediating Role of Self-relevant Emotions and Public
Risk Perception," Health Communication, pp. 1-10, 2020.
[13] Kementerian Kesihatan Malaysia, "COVID-19," Kementerian Kesihatan Malaysia, 2020.
[Online]. Available: https://www.facebook.com/kementeriankesihatanmalaysia/. [Accessed
27 February 2020].
[14] Noor Hisham Abdullah, "Noor Hisham Abdullah," Noor Hisham Abdullah, 2020. [Online].
Available: https://www.facebook.com/DGHisham/. [Accessed 27 February 2020].
[15] CoronaTracker Community , "CoronaTracker," CoronaTracker, 2020. [Online]. Available:
https://www.coronatracker.com/. [Accessed 28 February 2020].
[16] A. Rachah and D. F. M. Torres, "Analysis, simulation and optimal control of a SEIR model
for Ebola virus with demographic effects," Commun. Fac. Sac. Univ. Ank. Series A1, vol. 67,
no. 1, pp. 179-197, 2018.
[17] A. T. Porter, "A path-specific approach to SEIR modeling," Ph. D Thesis University of Iowa,
2012.
[18] M. Munikar, S. Shakya and A. Shrestha, "Fine-grained Sentiment Classification using
BERT," ArXiv, 2019.
[19] X. Dong and G. de Melo, "A Helping Hand: Transfer Learning for Deep Sentiment Analysis,"
in Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics,
2018.
[20] U. Upadhyay, "Knowledge Distillation," Medium, 2018. [Online]. Available:
https://medium.com/neuralmachine/knowle ... 241d7c2322. [Accessed 01
March 2020].
[21] John Hopkins University, "Coronavirus Map," John Hopkins University, 17 March 2020.
[Online]. Available: https://coronavirus.jhu.edu/map.html. [Accessed 17 March 2020].
[22] J. T. Wu, K. Leung, G. M. Leung, "Nowcasting and forecasting the potential domestic and
international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modelling
study," The Lancet, vol. 395, no. 10225, pp. 689-697, 2020.
[23] The Institute for Disease Modelling, "SEIR and SEIRs models," Institute for Disease
Modelling, 2019. [Online]. Available:
https://institutefordiseasemodeling.git ... -seir.html.
[Accessed 03 March 2020].
[24] Q. Li et al, "Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus–Infected
Pneumonia," The New England Journal of Medicine, pp. 1-9, 2020.
[25] R. Socher et al, "Recursive Deep Models for Semantic Compositionality Over a Sentiment
Treebank," in Conference on Empirical Methods in Natural Language Processing, 2013.
[26] T. Wolf et al, "HuggingFace's Transformers: State-of-the-art Natural Language Processing,"
ArXiv, vol. abs/1910.03771, 2019.
[27] CoronaTracker Analytics Team, "Corona Tracker Analytics," CoronaTracker, 03 March
2020. [Online]. Available: https://www.coronatracker.com/analytics. [Accessed 03 March
2020].
[28] V. Agafonkin, "Leaflet," Leaflet, 2010. [Online]. Available: https://leafletjs.com/. [Accessed
7 March 2020].
[29] SciPy Developers, "SciPy," SciPy, 2020. [Online]. Available: https://www.scipy.org/.
[Accessed 13 03 2020].
[30] J. An, K. Ahn and M. Song, "Text Mining Driven Content Analysis of Ebola on News Media
and Scientific Publications," Journal of the Korean Society for Library and Information
Science, vol. 50, no. 2, pp. 289-307, 2020.
[31] Kim, Y. Jeong, Y. Kim, K. Kang and M. Song, "Topic-based content and sentiment analysis
of Ebola virus on Twitter and in the news," Journal of Information Science, vol. 42, no. 6,
pp. 763-781, 2016.
[32] S. K. Khor, "The Politics of the Coronavirus Outbreak," Think Global Health, 24 January
2020. [Online]. Available: https://www.thinkglobalhealth.org/artic ... oronavirus
outbreak. [Accessed 04 March 2020].
[33] S. K. Khor, "Malaysia does not have a good record of transparency," The Star, 15 January
2020. [Online]. Available: https://www.thestar.com.my/opinion/columnists/vital
signs/2020/01/15/malaysia-does- not-have-a-good-record-of-transparency. [Accessed 04
March 2020].
[34] J. Passy, "Here's why the U. S. government's effort to contain the coronavirus outbreak was
never going to be successful.," Market Watch, 29 February 2020. [Online]. Available:
https://www.marketwatch.com/story/heres ... the-united
states-despite-government-efforts-to-contain-the-outbreak-2020-02-27. [Accessed 27
February 2020].
[35] D. E. Bloom, D. Cadarette and JP Sevilla, "New and resurgent infectious diseases can have
far-reaching economic repercussions," Finance and Development, vol. 55, no. 2, pp. 46-49,
2018.
[36] M. McCormick, "Coronavirus: ECB ready to take "targeted" action to address economic
impact of outbreak - as it happened," Financial Times, 2 March 2020. [Online]. Available:
https://www.ft.com/content/60f9e2ec-dd3 ... daa4dde1f8. [Accessed 04 March
2020].