Mining The Basic Reproduction Number (R0) Forecast For The Covid Outbreak
Loading...
Date
2022-07-25
Authors
Rajogoval, Illayakantthan
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
The popular method used to measure the spread of COVID-19 is by calculating
the Basic Reproduction Number, 𝑅𝑜. This number is a globally used metric to describe
the COVID-19 outbreak globally. Existing studies applied different models including
SIR, SIRD, SEIR, SPIR, MCMC, Statistical exponential growth, statistical likelihood
estimation and dynamic transmission models to evaluate COVID-19 𝑅𝑜. However,
there is no exact best model to ensure accurate Basic Reproduction Number, 𝑅𝑜 .
Besides, the essential attributes to return the best prediction model for COVID-19 𝑅𝑜
remains unclear. Therefore, this study aims to identify and evaluate the attributes and
parameters associated with the development of the Basic Reproduction Number, 𝑅𝑜
models, classify the data used in existing Basic Reproduction Number 𝑅𝑜 models,
develop a predictive classification model for the Basic Reproduction Number, 𝑅𝑜 and
to assess and enhance the accuracy of the Basic Reproduction Number, 𝑅𝑜 prediction.
Two case studies related to the COVID-19 outbreak in Malaysia and Malta were used.
The study attributes are mainly about daily cases, deaths, hospitalization, ICU
admission and Isolation. Waikato Environment for Knowledge Analysis (WEKA)
version 3.8 was adopted for data mining analysis at two levels of classification stages.
Data pre-processing procedures ensure outlier and extreme values and irrelevant
attributes are discarded. For classification analysis, J48, Naïve Bayes, Random Forest
and SMO were used on three different testing options: full training set, 10-fold cross
validation and 66% split to yield the percentage of classification accuracy. Accuracies
obtained for first-level classification ranged from 70.6349% to 100% for the Malaysia
dataset and 89.6266% to 100% for the Malta dataset.
viii
Meanwhile, for second-level classification, the accuracies ranged from
67.8751% to 100% for the Malaysia dataset while 58.9212% to 100% for the Malta
dataset. All the classification accuracies obtained were above the baseline accuracy. The
COVID-19 Basic Reproduction Number, 𝑅𝑜 a predictive model is developed using a
linear regression classification algorithm to predict the COVID-19 Basic Reproduction
Number, 𝑅𝑜based on the actual COVID-19 Basic Reproduction Number, 𝑅𝑜. The study
identified new cases, deaths, hospitalization, and ICU admission as important attributes
to derive accurate COVID-19 Basic Reproduction Number, 𝑅o.