Page 476 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

Geographically Weighted Regression Analysis for COVID-19 Twitter Data

*
Neda Kaffash Charandabi , Raziyeh badri, Nadia Tavakoli

Faculty of Geomatic, Marand Technical Faculty, University of Tabriz, Tabriz, Iran

*Corresponding author. E-mail addresses: n_kaffash@tabrizu.ac.ir

Abstract

At the time of writing, there were more than 148 million confirmed COVID-19 cases around the world, and the
virus's spread has already wreaked havoc on the citizens, resources, and economies of many countries. Globally,
social distancing steps such as travel bans, self-quarantines, and company closures are altering society's very
structure. Since people are being forced out of public areas, much of the discussion on these issues now takes
place on social networks such as Twitter. Communication platforms inspired by COVID-19 outbreak, and users
exchange various messages to keep each other informed. In this regard, the relationship between COVID-19
data and Twitter messages from people in various countries was investigated in this paper. More than 66 million
text tweets and 94 million location tweets were examined using the geographical weight regression method in
the first four months of the COVID-19 outbreak to find correlation between corona data (including mortality,
number of patients and recovered, and testers) and Twitter data (including users' tweets by post, geographical
location, and photo). The results indicate that COVID-19 data had a significant impact on people's tweets in
countries such as the United States, China, South Korea, and Japan. Furthermore, more than 61% of countries
have a low standard deviation in detecting these spatial auto-correlations.

Keywords: COVID-19, spatial autocorrelation, GWR, Twitter

Introduction
The first cases of COVID-19 were identified in late December 2019 in Wuhan, China, by the World Health
Organization (WHO) and the first deaths were reported in early 2020 [1]. Following that, it spread exponentially
around the world, affecting a large number of people [2]. Millions of people around the world are influenced by
preventative measures undertaken by governments [3, 4]. The most commonly used of these interventions,
social distancing, aims to reduce new infections by reducing physical interaction between people [5]. Many
companies have been forced to compel their employees to work from home as a result of social distancing
initiatives [6], school and college closures [7]. The discourse around COVID-19 has continued to evolve as
more social connections move online, with an increasing number of people turning to social media for both
information and company [9,10]. Platforms like Twitter, Instagram, and WhatsApp have become critical
components of the technical and social infrastructure that helps people to remain linked even in the face of such
crises. Social networks have the ability to reveal useful insights into human emotions. Investigation of tweets
could be useful, particularly during and after the COVID-19 pandemic, when the situation and people's reactions
are changing at a rapid pace. As a result, evaluating Twitter data may be critical in understanding people's
behavior and responses during the COVID-19 pandemic. Recent research [11]–[14] shows that using Twitter
data and analyzing human emotions can help forecast crimes, the stock market, election results, disaster