Temporal-Spatial Analysis based Public Sentiment about COVID-19 On Social Media
: A Case Study on Weibo In China

Student thesis: Master's ThesisMaster of Science


The outbreak of COVID-19 has blowout around the world, and many countries have seriously suffered from it. Analyzing how public sentiment changes over time is helpful for the policy makers to stabilize the society during this tough time, and meaningful for the scholars to study the social impact of coronavirus. Thus, this research taking Weibo (a social media platform in China) as an example, analyses public sentiment development from the perspective of space and time from January to June 2020.
Firstly, advanced web crawler technology based on Selenium is used to collect data from this platform. Finally, 60,000 pieces of posts uniformly distributing from Jan to June are collected.
Then, the sentiment analysis based Naïve Bayes is applied to assess emotional tendency of each post. A sentiment score ranging from 0 to 1 is given after sentiment analysis, in which 0 for negative and 1 for positive. Because of the limitation of the original model, three methods are introduced to further improve the accuracy and efficiency. At last, the accuracy successfully raises from 33.8% to 84.8%, and the running time halves
Thirdly, the temporal-spatial distribution and monthly variation of sentiment score are analysed. Spatial analysis is mainly global and local spatial autocorrelation analysis. Three methods, Moran’s I, Local Geary and Local Gi, to access local spatial correlation are applied. After comparison, the results of Moran’s I are more seasonable and reliable. According to the results, sentiment score in Mach and April has strong global correlation. And there is an obvious High-High type of spatial correlation in western area in March and a High-Low type of spatial correlation in western area in April. -Temporal-Spatial analysis is realised by Mann-Kendall Trend Test and visualised by hot spots map. Central and eastern area’s sentiment score show significant increasing trend during this period, while that of western area come back to original value after fluctuations.
Finally, some useful suggestion to manage the public sentiment is proposed to the Chinese government.
Date of Award23 Oct 2020
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorRita Borgo (Supervisor)


  • sentiment analysis
  • COVID-19
  • social platform
  • temporal-spatial
  • China

Cite this