Smart Methods to Deal with COVID-19 at University-Level Institutions Using Social Network Analysis Techniques

Round 1

Reviewer 1 Report

A review of this article provides a global view of the current global health crisis and highlights a major consequence of the pandemic caused by Covid-19. Thus, taking into account the objective pursued by the article, it is considered that the methodology proposed by the authors is adequate to the objective and the hypothesis proposed. Through this methodology, specifically, the method of application of SNA, clear and concise conclusions are reached where through social networking strategies, we can reduce the maximum number of infections. In addition, it has been observed that the citations are appropriate with the corresponding citation, as well as all this is framed in an updated framework.

Author Response

Thanks for your valuable comments.

Reviewer 2 Report

This manuscript by Malick et al. investigated “Smart Methods to Deal With Covid_19 at University Level Institutions using Social Network Analysis Techniques”. I have gone through the manuscript and the work is good but the authors failed to address the issues properly in all the sections of the manuscript. Overall, the whole manuscript is poorly written, the study is not well designed and executed, data properly analysed, methodology lacks of details and results not well discussed and discussion section is totally absent.

Abstract: This section is not well organized. There are many spelling mistakes such as variatns, virsu and some grammatical errors in this section. The authors should reduce the background study while they have to mention the methodology in few lines. Ultimately, the authors should rewrite this section.

Why did the authors write clustering coefficient and degree centrality in keywords section but these two words were totally absent in abstract?

Introduction: There are a lot of spelling mistakes and grammatical errors in this section. The authors did not organize this section. This section lacks of logical flow and also they did not mention the actual justification of this study. Therefore, they have to rewrite this section with some literature references.

Methodology and results

The authors should add a study map of the experimental site.

 In Figure 5, figure legend should be revised. All figures should be understandable and attracted to the readers.

Discussion

The authors just mentioned the results rather than the discussion section. Discussion must be included with this research.

Conclusion

The authors failed to write conclusion precisely. They have just described the methodology and results in this section. This section should be outcome based according to title and objectives.

The readers will be confused due to poor English writing. Therefore the manuscript should rewrite in an international standard through a native English speaker.

Author Response

Response to the comments made by the reviewers

First of all, the authors of the paper would like to thank the editor and reviewers for their helpful and constructive comments and suggestions. We have carefully modified the paper according to their comments. The detailed information is shown as follows:

Reviewer#2, Concern # 1:  This manuscript by Malick et al. investigated "Smart Methods to Deal With Covid_19 at University Level Institutions using Social Network Analysis Techniques". I have gone through the manuscript and the work is good but the authors failed to address the issues properly in all the sections of the manuscript. Overall, the whole manuscript is poorly written, the study is not well designed and executed, data properly analysed, methodology lacks of details and results not well discussed and discussion section is totally absent.

Author response:  Thank you for providing us with key insights about various sections, including a discussion of results. We have updated all sections and provided the details in response to the respective comment. We have added a new section named "Discussion and Limitations".

Author action: We updated the manuscript by revising the following sections; Abstract, Introduction, Research Methodology, Data Collection,Network Analysis, Results. Added a new section named "Discussion and Limitations"

Reviewer#2, Concern # 2: Abstract: This section is not well organized. There are many spelling mistakes such as variatns, virsu and some grammatical errors in this section. The authors should reduce the background study while they have to mention the methodology in few lines. Ultimately, the authors should rewrite this section.

Why did the authors write clustering coefficient and degree centrality in keywords section but these two words were totally absent in abstract?

Author response:  Thanks for your valuable comments. The Abstract has significantly changed and updated according to the journal standard. We have removed the keywords clustering coefficient and degree centrality from the keyword sections.

Author action: Grammatical mistakes are removed, and thorough spelling and grammatical checks are made. Now the abstract has no spelling or grammar issues. The changes are highlighted in the supported document attached.

Reviewer#2, Concern # 3:  Introduction: There are a lot of spelling mistakes and grammatical errors in this section. The authors did not organize this section. This section lacks of logical flow and also they did not mention the actual justification of this study. Therefore, they have to rewrite this section with some literature references.

Author response:  Thanks for your valuable comments. According to the reviewer's comment, a specific paragraph has been added in the introduction, citing the current state of the art. The newly added paragraph (citing the current state of the art) can be found in the author action section. Following that, contributions are highlighted by the end of the introduction section and can be found in the author action section. Moreover, a section is added in the introduction, referring to the overall organization of the paper and logical flow. The added sentences can be found in the following section.

Author action: We updated the manuscript through thorough spelling and grammatical mistakes.

The following paragraphs are added in the introduction section;

"Because of the unprecedented situation caused by the pandemic, several studies are proposed to mitigate the adverse effects of the complete lockdown at universities. A holistic approach is presented that recommends institutions to early identify the infected ones by speeding up the detection process, which will reduce the rate of spread of Covid-19 in respective groups [5]. A simulation-based study evaluated the effectiveness of the spread of Covid-19 viruses by limiting human mobility. The study considered group-to-group interactions instead of considering individuals as spreaders among the groups [6]. Considering e-learning as an alternative to the physical learning environment left as the only option, however, improving group-based learning methodology is proposed by [7] to reduce the spread faster. The Covid-19 era has passed. However, few interventions have been made to identify smart methods for keeping the universities running while reducing the risk of spread in the meantime. A strategy is proposed to identify the critical nodes first in order to prioritize the process of vaccination with the expectation of minimizing the spread over the network [8]. This study is closest to our method while considering the human interactions in the form of networks and then vaccinating the key figures, having the most of the risk of spread accordingly. After an exhaustive literature search, it can be stated that social network analysis-based elimination of key students in universities is not studied well and demands detailed analysis. "

Following lines are added in the next paragraph:

"According to the social network (SN) theories, human interaction tends to form networks. These networks do hold several attributes that can be referred to as complex network-based technically. In most real-world networks, structural patterns exist that can be used to model the rate of spread of information, virus, technology adoption, and many more. "

The following is added at the end of the introduction section;

"The contributions of this paper are as follows:

1. A novel and smart method is proposed and experienced to identify the interaction networks among university-level students along with the identification of critical nodes.
2. Effectiveness of network modeling techniques is presented for the student's interaction networks.
3. The study shows that the removal of a few critical nodes from our methods, i.e., the students, will significantly drop the rate of spread of the potential virus among the entire network. Higher education institutions can use the method to avoid a complete lockdown in case of any pandemic in the future. "

The overall organization of the paper:

"The rest of the paper is organized as follows. Section 2 highlights the discussion over research methodology, which is followed by the data collection process described in Section 3. The network analysis is done in Section 4. The results and discussion of those results are formulated in Sections 5 and 6, respectively. The research methodology highlighted the methods of modeling the SNA along with the comparison among them. Furthermore,  the pragmatics of employing SNA techniques over the students' interaction networks are presented in detail under the context of virus spread on established spread models. "

Reviewer#2, Concern # 4: Methodology and results

The authors should add a study map of the experimental site.

In Figure 5, figure legend should be revised. All figures should be understandable and attracted to the readers.

Author response:  Thanks for your valuable comments. A thorough checkup is made for spelling and grammatical mistakes, and updated the original manuscript. A new figure is added in the methodology section to present the flow of the study. All the figure captions are rewritten and made self-explanatory. Along with the figure captions, captions with the tables are also rewritten and presented below.

Author action: We updated the manuscript by improving the grammar and spelling mistakes.Figure 1 with the caption is added in the updated manuscript.

Figure 1 Research Methodology: This represents the flow of the study and presents the stages of the study, including surveying students, modeling social networks, observing key attributes at the network level, applying the SIR model, and role of node removal over different types of networks. 

Following Captions of Figures 2-9 are updated in the revised manuscript.

Figure 2 For a better understanding of the modeled networks, two generated networks are presented: Watts Strogatz (left) and Erdos-Renyi (right) based models for the same  class BCS-4F. Both network models are distinct in terms of centrality measure and clustering coefficient. The Erdos-Renyi model is dense with the absence of highly centric nodes, and Watts models do contain the highly centric nodes in the generated model. 

Figure 3. The network generated by the sequential method is different from Erdo-Renyi, along with some similarities with the Watts model. Network generated using Sequential model for section BCS-4F.

Figure 4. Density comparison between networks generated through Sequential, Watts Strogatz, and Erdos modeling techniques. This shows that the networks generated through the Watts model are dense in nature in comparison to the other two methods.

Figure 5. The clustering coefficient is a strong measure to interpret the nature of the complex real-world network. A comparison between clustering coefficients is presented based on the used models, including Sequential, Watts Strogatz, Erdos-Renyi

 Figure 6 Impact of node removal on density. It is shown in the figure that the removal of nodes has a significant difference on several network attributes, including density, Betweenness, avg centrality, closeness centrality, and PageRank centrality. The impact of node removal is observed over the networks generated through the Sequential method and presented results in comparison with networks generated through a random model.

Figure 7. The impact of node removal on the clustering coefficient is evaluated among the networks generated through the Sequential method and Random method. The trend shows that the removal of significant nodes (on the two types of networks) results in different cohesion.

Figure 8. The impact of node removal on average path length is evaluated among the networks generated through the Sequential method and Random method. The trend shows that the removal of significant nodes (on the two types of networks) has different results. The random model-based networks have little impact on the average path length after removing significant nodes. 

Figure 9. Virus transmission on the network using different strategies. The rate of spread of the virus is predicted after removing nodes through multiple criteria, including degree, betweenness, random, PageRank, and Closeness centrality. The results show that the removal of  highly centric nodes tends to reduce the spread of the virus over generated models. Betweenness centrality-based removal was found second best closer to the Closeness centrality measure.

The captions of the tables are updated.

Table 1. The table shows the distribution of edges among the nodes based on the used several models. Network properties are also presented in the table to observe the difference between the result of each model. Erdos-Renyi model has a higher density on average with a low clustering coefficient with respect to the two other models.

Table 2. The table shows the distribution of edges among the nodes based on the used several models. in the population. Network properties are also presented in the table to observe the difference between the result of each model. Erdos-Renyi model has a higher density on average with a low clustering coefficient with respect to the two other models.

Reviewer#2, Concern # 5: Discussion The authors just mentioned the results rather than the discussion section. Discussion must be included with this research.

Author response:  Thanks for your valuable comments. A separate section titled; "Discussion and Limitations" is added to the revised manuscript. 

Author action: We updated the manuscript by adding the following section under the heading of discussion and limitations.

"This study has demonstrated a strategy that can be used to deal with the emergency state at educational institutions in COVID-19-hit cities. A detailed survey was conducted at a university to identify the homogeneity among the network structure at inter and intra-class interactions. Initially, the data was collected in batches for generating Networks using different models like the Watts-Strogatz model [23], the Erdos-Renyi model [24], and the Sequential Algorithm model [25]. Network centrality measures were computed, including Average Degree, Density, and Clustering Coefficient. The generated networks were then evaluated regarding resilience towards rapidly spreading the virus based on the found interactions.

The node removal method was employed to evaluate the resilience of the networks. The nodes were removed from the network by using popular, influential node identification strategies; based on degree centrality, betweenness centrality, closeness centrality, page rank, and based on random removal. Considering each centrality measure, the top-scored nodes were removed, and then the SIR model was applied to analyze the transmission of the contagions. Iteratively, all the generated networks are evaluated, and it is found that highly centric node removal reduces the spread compared to the randomly selected node removal. In the presence of complex network structures having highly centric nodes, that tends to spread the virus among the neighbors rapidly and reduces the overall network diameter. That can be interpreted in the prevalence of complex network attributes in student interaction networks.

These results will be helpful in managing the classes during the serge of any possible COVID-19 variant by isolating the influential nodes or prioritizing complete vaccination courses. From the result, we can observe that nodes selected for vaccination from centrality measures decrease the density and other cohesion metrics compared to randomly selected nodes. The spread of the virus can be reduced by simply choosing the top network contributors and isolating or immunizing them. The overall positive rate of COVID-19 can be reduced, and its dissemination can be controlled by choosing and immunizing people with high-centric values. By using such techniques, universities can avoid total lockdown, or without achieving "complete lockdown," they can continue their academic endeavors. The schools have different social structures, particularly in South East Asia. Therefore, it is mentioned repeatedly that the results only apply to the universities unless applied and evaluated at school-level students."

Reviewer#2, Concern # 6: Conclusion

The authors failed to write conclusion precisely. They have just described the methodology and results in this section. This section should be outcome based according to title and objectives.

The readers will be confused due to poor English writing. Therefore, the manuscript should rewrite in an international standard through a native English speaker.

Author response:  Thanks for your valuable comments. We realize that the conclusion must be rewritten, which is presented in the following section.

Author action: The conclusion is rewritten and added to the conclusion section. It can be found below:

"The global pandemic spread greatly affected social interactions independent of age, location, and race. The educational institutions received the worst impact because of the complete lockdown. The presented study offered an alternate solution to the complete lockdown, particularly targeting the higher education institutions in Southeast Asian countries. The study identified the underlying structural interaction pattern among university students evaluated through social network analysis techniques. The social network analysis techniques focused on the centrality measures to identify key individuals among the predicted students' networks. The interaction networks are analyzed in terms of the spread of disease with the help of spread models, which showed that removing a few individuals decreased the spread rate among the contagions. The results offer an opportunity for institutes of higher education to continue educational activities with the elimination of a few students. The strategy can be considered a smart method to reduce the spread of the disease while continuing educational activities without complete lockdowns.

The present study can also benefit offices and other educational institutions by implementing similar experiments. This strategy will allow a continuous availability of professional services and educational activities, even during any potential pandemic in the future."

Author Response File: Author Response.pdf

Reviewer 3 Report

Very actual topic!

Too much keywords.

Author Response

Thanks very much for your valuable comments, we have updated the keywords. 

Updated keywords in the revised manuscript.

Keywords: COVID-19; Social Network Analysis (SNA), smart vaccination; smart lockdown; university education; network-based strategy. 

Round 2

Reviewer 2 Report

I have gone through this manuscript by Malick et al. entitled “Smart Methods to Deal With Covid_19 at University Level Institutions using Social Network Analysis Techniques”. Now the authors have improved significantly this manuscript and this is much better than the previous one. But the discussion section should improve. I appreciate all of the authors of this research work for their hard labor.

Minor revision: The discussion section should improve.

Best of luck.

Author Response

Comment 1: I have gone through this manuscript by Malick et al. entitled “Smart Methods to Deal With Covid_19 at University Level Institutions using Social Network Analysis Techniques”. Now the authors have improved significantly this manuscript and this is much better than the previous one. But the discussion section should improve. I appreciate all of the authors of this research work for their hard labor.

Author response:   We are thankful to the honorable reviewer for encouraging comments. 

Comment 2: Minor revision: The discussion section should improve.

Author response:  Thanks for your valuable comments. The discussion section has significantly changed and been updated.

Author action:  Following highlighted text is updated in the “Discussion and Limitations” section.

Discussion and Limitations

This study has demonstrated a strategy that can be used to deal with the emergency state at educational institutions in COVID-19-hit cities. A detailed survey was conducted at a university to identify the homogeneity among the network structure at inter and intra-class interactions. Initially, the data was collected in batches for generating Networks using different models like the Watts-Strogatz model [23], the Erdos-Renyi model [24], and the Sequential Algorithm model [25]. The modeled networks represent the social interaction networks of each class. The assessment of social networks is widely performed based on multiple measures, including; centrality, modularity, and resilience. The centrality measure refers to the structural representation of the network regarding the presence of highly centric nodes that distinguishes a random network from a classical complex network. The modularity represents the overall topological connectedness among the subnetworks of the networks. A network's resilience indicates a network's ability to hold its properties after removing certain nodes. In the present study, network centrality measures were computed, including Average Degree, Density, and Clustering Coefficient. Based on the found interactions, the generated networks were then evaluated regarding resilience towards rapidly spreading the virus.

The predicted social networks of students' interaction are evaluated to compute the resilience of these networks. The notion behind evaluating resilience is to identify the networks' structural sensitivity over removing fewer nodes. While considering the predicted students' network as complex networks, it is hypothesized that removing a few 'selected' nodes would have more impact on the networks than removing randomly picked nodes. The node removal method was employed to evaluate the resilience of the networks. The nodes were removed from the network by using popular, influential node identification strategies; based on degree centrality, betweenness centrality, closeness centrality, page rank, and based on random removal. Considering each centrality measure, the top-scored nodes were removed. The impact of the removal of prospective influential nodes in comparison to the randomly selected nodes is first observed over the network properties, including; the average path length, density, and clustering coefficient of each network. To address the core problem of restricting the potential spread of the virus over the network, we further investigated the spread over the resultant networks. The Suspect Infect and Recover SIR model was used to evaluate, and then the SIR model was applied to analyze the transmission of the contagions. Iteratively, all the generated networks are evaluated, and it is found that highly centric node removal reduces the spread compared to the randomly selected node removal. In the presence of complex network structures having highly centric nodes, that tends to spread the virus among the neighbors rapidly and reduces the overall network diameter. That can be interpreted in the prevalence of complex network attributes in student interaction networks.

The present study limited its scope to understanding the intra-class interaction patterns that can be considered strong ties in the literature on social networks. However, one can investigate the impact of the long ties while viewing the entire university as a single network to model the spread of long-range connections, e.g., through members of playing groups or drama society. The interacting students may not be highly centric within the class but can transmit the virus in communities at longer distances. Furthermore, the area of complex networks has recently observed advances in mesoscopic centrality awareness node significance. It refers to the idea that community structures may possess significant information regarding highly centric individuals within the community but do not receive attention from centrality score mining algorithms at the network level. This information may open new avenues for scientists to examine the discussed problem differently. 

These results will help to manage the classes during the serge of any possible COVID-19 variant by isolating the influential nodes or prioritizing complete vaccination courses. From the result, we can observe that nodes selected for vaccination from centrality measures decrease the density and other cohesion metrics compared to randomly selected nodes. The spread of the virus can be reduced by simply choosing the top network contributors and isolating or immunizing them. The overall positive rate of COVID-19 can be reduced, and its dissemination can be controlled by choosing and immunizing people with high-centric values. By using such techniques, universities can avoid total lockdown, or without achieving "complete lockdown," they can continue their academic endeavors. The schools have different social structures, particularly in South East Asia. Therefore, it is mentioned repeatedly that the results only apply to the universities unless applied and evaluated at school-level students.