Relationship between user behaviors and number of COVID-19 positives interpreted from Twitter data #12

Date: 2021.11.16
Researcher: Masashi Toyoda
Organization: Institute of Industrial Science, the University of Tokyo

Analysis of risk behavior-related Tweets

・Tweets related to actions of high infection risk were extracted and compared with positive COVID-19 cases. Tweets including the following keywords were subject to analysis:
　-Drinking session, End-of-year Party, Christmas Party, New Year Party (Respectively 飲み会、忘年会、クリスマスパーティ、新年会 in Japanese)
　-Karaoke (カラオケ in Japanese)
　-Fitness center (ジム in Japanese)
　-Hanami (“Flower-viewing”, 花見 in Japanese)
　-Barbecue (バーベキュー in Japanese)
・The situations within Tokyo and Osaka regions were compared with each other, using location information from Twitter user profiles.

Summary

・Excursion-related tweets nationwide are dwindling after the fluctuations related to the termination of a state of emergency subsided.
　-Karaoke numbers rose sharply after the termination, followed by a slow decline; the figures currently remain similar to that of July this year.
　-Drinking session numbers continued rising after the termination, but is currently staying at the levels of early October last year.
　-Numbers in Tokyo and Osaka follow similar patterns. Osaka is seeing a relatively higher number of Karaoke-related tweets.
　-In Hokkaido, Karaoke-related tweets increased rapidly after the termination, and has remained in that level since. Drinking session numbers are continuing to rise as well.
・Negative correlation between the number of tweets related to risk-related activities and the number of positive COVID-19 cases.
　-Tokyo seems to be more sensitive to the number positive COVID-19 cases, compared to Osaka.