Project Review-3

MAJOR PROJECT WORK
B. Tech - VIII SEMESTER

[2020 – 2021]
TITLE: Machine Learning Based Probabilistic Evaluation of COVID-19
Abstract
The COVID 19 pandemic has affected the world very badly as the
number of cases is not declining any day. The growth and the
mutation of the virus in a different form have created a lot of
concern among the researchers working on it. This project enables
the prioritization of tests by taking the symptoms from a person and
feeding the input data to it which in turn gives an output of
estimation of how much the person is infected with COVID 19.
Introduction
 COVID-19 is a viral infectious disease in which Wuhan was the initial epicenter and the
maximum confirmed cases worldwide as of 8 November 2020. Coronavirus 2, also
known as SARS-CoV-2 Extreme Acute Respiratory Syndrome, is the virus which
caused the COVID-19 pandemic. This virus spreads rapidly when the infected person is
in close contact. The lower concentrations respiratory droplets when an infected person
coughs or sneezes, which may not be visible to the naked eye.
 These droplets also may originate from saliva, which can be inhaled through the mouth
or nose into a healthy person's lungs, thus spreading the disease from one person to
another. If they touch their eyes, nose or mouth after they touch any contaminated
objects, people may also get sick.
 Old people with high blood pressure, heart attacks and diabetes have a serious risk of
COVID-19 infection. Aggregated data on the symptoms of other infected patients would
play an enormous role in this study in order to speed up the early detection of COVID-
19 infected individuals.
Existing System
• The COVID-19 tracking applications are being developed rapidly by

many countries around the world. There is a mobile application that
has been developed by the Indian government known as "Aarogya
Setu".
• This app includes features of tracking a person that is infected by
COVID. It has GPS and Bluetooth technologies to send alerts when
they come close to any infected person.
• It uses data science concepts like classification and clustering to
analyze the spread in India. It has an artificial intelligence feature to
detect patients of COVID-19. It is a very useful application that
introduces the tracking and safety of people.
Proposed System
• A more specific and accurate estimation of the disease can help bring
alertness to a person. This application gives a probabilistic result set
with 70% accuracy based on the nationwide data.
• To be specific, the option to check doctors nearby and get an
appointment or the needed medication at the moment can add more
credibility to the application. It can use more datasets and resources in
updating the limited symptoms to be used in estimating the disease.
• Classification of the data sources of medication for each symptom and
self-isolation tips can be an add-on. Step-by-step guidance of self-
isolation and medication details, course, and timings check can also
be developed and updated in the application.
Problem Statement
This project enables the prioritization of tests by taking the symptoms from a
person and feeding the input data to it which in turn gives an output of estimation
of how much the person is infected with COVID19.
Objective
 To predict the likelyhood of a person whether effected with covid or not
and by how much percentage did a person get effected.
 The objective of this study is to conduct the literature review on the role of
machine learning as a determined technology to fight against the COVID-
19 crisis in the sector of diagnosis and disease analysis as well as
estimation.
 We present a methodology that combines disease which estimates from digital
traces, via interpretable machine learning methodologies, to analyse the COVID-

19 activity in provinces in real time.
Literature Survey
Authors Names Title of the Techniques/Met Data Used Accuracy and
paper hods/Algorithm Observation
used
Wolfram Research "Patient Medical Random Forest Wolfram Data No of cases
Data for Novel Classifier https://doi.org/10.24 increasing
Coronavirus COVID- 097/wolfram.11224.d
19" ata
Moritz Kraemer Epidemiological data Naive Bayes WHO, U.S. CDC, Vaccine required
from the 2019 nCoV ECDC, China CDC
outbreak: early (CCDC), NHC and
descriptions DXY.
Dong,E., Du, H An interactive web- Real Time Covid-19 Resource Update the cases
& Gardner, L.An based to track Centre,Twitter throughout the day
COVID-19 in real time feeds,online news when new cases are
services identified
Menni, C. et al Real-Time tracking of Real time WHO,news services Tracking of potential
self-reported Nat. Med symptoms
symptoms to predict
potential COVID-19
Yazeed Zoabi, Shira Machine learning- Setting and study clinical Health system
Deri-Rozov & Noam based prediction of data symptoms,binary response to future
Shomron COVID-19 diagnosis indication epidemic waves of this
based on symptoms disease and other
respiratory viruses
Methodology
 First of all, the COVID-19 Symptoms dataset is uploaded and the attributes are
separated into independent and dependent variables, then the dataset is split into two
parts - one being the train data and the other being the test data by which we can
predict. The former being the larger part we process it through different machine
learning algorithms to get the best possible output, and we observe that by using the
Random Forest Classifier algorithm we get the best possible accuracy. The
performance of the Random Forest Classifier was the highest as compared to other
algorithms.
 The output variable is the target variable and all the other attributes of the symptoms
and individual’s necessary details are fed into it and the desired result is shown.
 After the training of the model is complete, then the later dataset can be used for
predicting the presence of COVID 19 in a particular individual. The Covid Output is
the final result that we need to predict from our study. After the utilization of different
machine learning algorithms, we construct a model that can be used to predict the
patient’s final Covid Output.
Block-Diagram
Algorithms
•Logistic Regression
•Decision Tress Classifier
•Random Forest Classifier
•Naïve Bayes
Data Description
Sample data
Source of Dataset : Wolfram Data Repository
Description of each parameter
Results and Discussion
Preliminary Results
Precision: It is defined as the number of true positives divided by
the total count of false positives and true positives.
F1 Score: It is defined as the weighted average of the precision and
recall.
Various classification algorithms are applied and the COVID19
Symptoms dataset is provided for the performance analysis of
different models.
Result
Input to Output clear representation
Conclusion
• At present the information about the symptoms to COVID-19 infected outcome
is present at a very small scale while the future work will be able to draw a better
understanding between the symptoms and the infection in a greater detail as the
contribution of adding new data will be done.
• This project gives an insight of the different symptoms linked to COVID-19 and
proposed a probabilistic classification of getting infected. The work concludes
that men have a higher tendency to get infected having common symptoms as
fever and cough. To choose the best algorithm to train the dataset, statistical
analysis is carried out.
• Random Forest Classifier Algorithm is superior to other models in comparison to

various performance factors such as accuracy, AUC Score, F1 Score and Recall.
The approximate mean accuracy score using these algorithms was found to be
78%.
References
List of References – used in Literature survey (Appropriate to Project
problem and latest)
Wolfram Research, "Patient Medical Data for Novel Coronavirus COVID-19" from the Wolfram Data
Repository (2020) https://doi.org/10.24097/wolfram.11224.data
What are the symptoms of COVID-19?,https://www.who.int/emergencies/diseases/novelcoronavirus-
2019/question-and-answers-hub/q-adetail/q-a-coronaviruses#:~:text=symptoms
Coronavirus disease 2019 (COVID-19),
https://www.mayoclinic.org/diseasesconditions/coronavirus/diagnosis-
treatment/drc20479976
Dong, E., Du, H. & Gardner, L. An interactive web-based dashboard to track COVID-19 in real time. Lancet
Infect https://doi.org/10.1016/S1473-3099(20)
Punn, N. S., Sonbhadra, S. K. & Agarwal, S. COVID-19 Epidemic Analysis using Machine Learning and Deep
Learning Algorithms. medRxiv, https://doi.org/10.1101/2020.04.08.20057679 (2020).

COVID-19-Government Data. https://data.gov.il/dataset/covid-19 (2020).
Zoabi, Y., Deri-Rozov, S. & Shomron, N. Machine learning-based prediction of COVID-19 diagnosis based on
symptoms. npj Digit. Med. 4, 3 (2021). https://doi.org/10.1038/s41746-020-00372-6

Menni, C. et al. Real-time tracking of self-reported symptoms to predict potential COVID-19. Nat. Med. 26,
1037–1040 (2020).
Whittington, A. M. et al. Coronavirus: rolling out community testing for COVID-19 in the NHS. BMJ Opinion
https://blogs.bmj.com/bmj/2020/02/17/coronavirus-rolling-out-community-testing-for-covid-19-in-the-nhs/
(2020).

Project Review-3

Uploaded by

Copyright:

Available Formats

Project Review-3

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Project Review-3

Uploaded by

Copyright:

Available Formats

MAJOR PROJECT WORK

B. Tech - VIII SEMESTER

• The COVID-19 tracking applications are being developed rapidly by

 We present a methodology that combines disease which estimates from digital

traces, via interpretable machine learning methodologies, to analyse the COVID-

•Decision Tress Classifier

•Random Forest Classifier

• Random Forest Classifier Algorithm is superior to other models in comparison to

Learning Algorithms. medRxiv, https://doi.org/10.1101/2020.04.08.20057679 (2020).

symptoms. npj Digit. Med. 4, 3 (2021). https://doi.org/10.1038/s41746-020-00372-6

You might also like