Identifying COVID-19 Hotspots using Bipartite Network Approach

• Main Author: Hong, Boon Hao
• Format: Thesis
• Language: English; English; English; English
• Published: University of Malaysia Sarawak 2023
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://ir.unimas.my/id/eprint/41679/5/MasterSci.%20Thesis_Hong%20Boon%20Hao%20-%20%2024%20pages.pdf; http://ir.unimas.my/id/eprint/41679/6/MasterSci.%20Thesis_Hong%20Boon%20Hao_fulltext.pdf; http://ir.unimas.my/id/eprint/41679/8/MSc.%20Thesis_Hong%20Boon%20Hao_%20supervisor%20approval%20form%202.pdf; http://ir.unimas.my/id/eprint/41679/9/MSc.%20Thesis_Hong%20Boon%20Hao_%20supervisor%20approval%20form%201.pdf; http://ir.unimas.my/id/eprint/41679/
• Institution: Universiti Malaysia Sarawak

The COVID-19 pandemic has affected countries worldwide, causing major disruptions in both health and economic systems. While lockdowns have been effective in controlling the spread of the virus, their negative impact on the economy has prompted the need for alternative, cost-balanced control measures. Contact tracing has emerged as a promising solution in identifying community outbreaks of COVID-19. To improve the efficacy of contact tracing, this study aimed to formulate a contact network model for COVID-19 transmission. Conventional approaches were found to be inadequate in modelling the transmission of COVID-19, particularly in identifying the source of infection. To address this, the study utilized a bipartite network modelling approach to account for the heterogeneity of transmission routes, human hosts, and visited locations. The human host and visited location were identified as the two discrete entities in the research scenario. Using data from the Bintulu Health Office's contact tracing investigation forms, six network models were formulated. The link weight between the human host and location nodes was quantified using the summation rule, taking into consideration various factors such as environmental properties, building characteristics, human and pathogen characteristics, and transmission modes. The location and human nodes were then ranked using a web-based search algorithm based on their respective ranking values. The results of the study showed that the bipartite network modelling approach was successful in formulating the contact network model. Verification analysis revealed a root mean square error of 0.0002244 and 0.001419 for the location and human nodes, respectively, which were well within the threshold value of 0.05. The ranking between the target and validated models was found to have strong similarity with a good Spearman’s rank correlation coefficient of above 0.70 (p < 0.001), indicating a high degree of relevance in improving contact tracing for COVID-19. The study also found that all parameters used in the model were relatively significant, and that the model had the ability to predict potential hotspots with 90% accuracy within a 600m radius for the subsequent week. These findings highlight the potential of the bipartite network modeling approach in improving contact tracing for COVID-19 and reducing the spread of the virus in high-risk areas.