ScholarWorks@성균관대학교

초록

The coronavirus disease 2019 (COVID-19) pandemic, which emerged in early 2020, has significantly affected global health and socioeconomic conditions. With significant fluctuations in case numbers, especially during critical events such as changes in the definition of the disease and diagnostic criteria, the need for accurate and early prediction of case numbers has become imperative. This study investigated the prediction of COVID-19 cases during the initial stages by merging (i) physical and behavioral data found in contact graphs and (ii) social media data. Recognizing the research gap in these multiple data modalities, we explored their potential synergy and the ensuing interplay between them to enhance the accuracy of outbreak prediction. To improve the predictive performance, we leveraged a cross-modal attention mechanism and investigated various fusion techniques. The major contributions of this study include the innovative use of multimodal data features and the development of a comprehensive methodology that integrates these features, allowing for a detailed analysis of the early dynamics of the COVID19 pandemic. The paper concludes with a thorough discussion of the experimental results and outlines directions for future research on pandemic prediction modeling.

키워드