A Framework for Crowd Management during COVID-19 with Artificial Intelligence

Round 1

Reviewer 1 Report

Overall work is good. Some comments are:

1) Author should write algorithm using formal notations.

2) There is no need to write code in the paper. Write procedures or algorithms.

3) There is heading 3.4.1 on page 10 and then on Page 11 there is heading 3.3.2. Check all headings.

 

Author Response

Indeed we are very thankful to you for making very constructive comments and pointing out shortcomings. In view of your comments, the manuscript has been revised thoroughly. The significant changes are as follows.

1. Algorithm is rewritten by using formal notation, and removing anomalies
2. Figure 5 is redrawn in the vector graphic format, which has significantly enhanced the resolution. We have removed the anomalies in the diagram as were pointed out but the second reviewer.
3. A through revision for language has been undertaken by a native English speaker, as a result of which all mistakes are removed and numerous improvements have been made to the manuscript.
4. Arguments and discussion of findings have been revised. In particular, parts in the Introduction and Conclusion sections have been improved. These modifications have improved the research design, questions, hypotheses and methods.
5. Abstract has been modified.
6. Important changes, except the Algorithm and Diagram, are highlighted

Author Response File: Author Response.pdf

Reviewer 2 Report

The quality of the article does not meet the level of the journal.

Some of the problems found are presented below for examples.

 

"Currently we are passing through the COVID-19 pandemic, a disease caused by coronavirus [1-4]."

This sentence is trivial, does not need to cite 4 articles.

 

"These restrictions absolutely essential and ..."

should be "These restriction are absolutely essential and ..."

 

"Global efforts to check the spread of the virus has so far received mixed success."

decide singular or plural

 

"The World Health Organisation (WHO) has approved several vaccines for emergency use but their effectiveness, especially against Delta variant, is questionable [6-7]."

The conclusion is too simple and too straightforward, does not meet the results of the two scientific articles.

 

"There is very little success in the treatment of the COVID-19 disease [8-9]."

The conclusion is too simple and too straightforward, does not meet the results of the two cited scientific articles.

 

"However, certain measures, recommended by the  WHO, have  been proven to be effective to contain the  COVID-19  virus,  and  are  universally agreed upon."

Not clear, what do certain measures contain. Surely not the virus.

 

"For, a number of home-grown apps are provided, which also display the status of health and vaccination of the pilgrims at the entry level."
This sentence is not clear enough.

 

"... was open only for local pilgrims but it  still attracted  millions of pilgrims and pandemic restrictions were not observed."
The restriction "allowing only local pilgrims" was probably a pandemic restriction.

 

"One  such  even  was"

even -> event

 

"The framework is supported with an Algorithm,."

Extra comma before end.

 

"We also provide validation of our algorithm.."

Extra point at end of sentence.

 

"Ore design takes into consideration ..."

Ore->Our

 

"...  aspect of social behaviour in a crowded even during a pandemic are discussed."

even -> event

 

"... notable them are ..."

bad grammar

 

"Hajj and Umrah [23] are very well known annual events which in normal times about ten million people permitted to perform these pilgrimages ..."

bad grammar

 

 

Figure 5. A Framework of Crowd Management during Pandemic.

"CVID-19 Kit" - missing "O"

The image is of bad quality.

Resolution should be higher, or the image should be vector graphic format.

Nouns tend to start with capital, which is strange, few start with minuscule.

There are logical problems with the figure, eg.: "Requirements for participation permit" is attached to the arrow with label "Apply for event permit"; "passage through thermal screening and sanitization tunnel" is not done by the "event management committee",

 

"By means of sensing components or event based decisions the data being  captured  and  applied  with  semantic  modelling  and  reasoning  through  machine learning model."

being -> is being

 

"... same will participates ..."

participates -> participate

 

The algorithm for the proposed model has bad layout and many other problems, e.g.:

function Smart_Application (input Event)
Event e= new Event ()
e<-{event_communication, event_bulletin, event_advertisement}

new instance in variable overwritten in next line?

"Step-4: function Event_Approaval ()"
typo in function name

"dataset<-visualize the training set results" - assignment operation overwrites dataset with visualization?
some functions have #Start and #End remarks, others do not

"function Event_Aprroval ()"
typo in function name

in function AI_Driven_Smart_Event_API ()
decision_set<-function Discover_Decisions ()
     if(decision_set==success)
the function call does returns a decision support data structure of predictions and dataset which cannot be compared with "success" 

Author Response

Indeed we are very thankful to you for making very constructive comments and pointing out shortcomings. In view of your comments, the manuscript has been revised thoroughly. The significant changes are as follows.

1. Algorithm is rewritten by using formal notation, and removing anomalies
2. Figure 5 is redrawn in the vector graphic format, which has significantly enhanced the resolution. We have removed the anomalies in the diagram as were pointed out but the second reviewer.
3. A through revision for language has been undertaken by a native English speaker, as a result of which all mistakes are removed and numerous improvements have been made to the manuscript.
4. Arguments and discussion of findings have been revised. In particular, parts in the Introduction and Conclusion sections have been improved. These modifications have improved the research design, questions, hypotheses and methods.
5. Abstract has been modified.
6. Important changes, except the Algorithm and Diagram, are highlighted.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The quality of the article has improved a lot, however, it still has fundamental issues.

See the followings for some important problems.

 

There are inappropriate (self)citations, e.g.:

"Currently we are passing through the COVID-19 pandemic, a disease caused by a coronavirus [1]."
The cited article is not an original source to be referenced for this fact.

"The World Health Organisation (WHO) has approved several vaccines for emergency use but their effectiveness, especially against the Delta variant, is questionable [3-4]."
Exactly what is questionable? The vaccines have proven (measured) efficiency for which no proper reference is included and considered, so the statement is biased.

"There has been very little success in the treatment of the COVID-19 disease [5-6]."
There are many reliable sources about the efficiency of treatment methods of COVID-19, instead of them a less cited source ([5]) is used here, and a more cited article ([6]) which does not present the conclusion that "There has been very little success in the treatment of the COVID-19 disease".

In Figure 5.

"Event Advertisement"

but

"Information bulletin" and "Event communication",

so the usage of Capitals is not consequent, as I mentioned in review1 already.

"Gathers Information/Initiate Required Action"

Gathers -> Gather

One of the white boxes does not have title as the other two boxes (it is not required, it's just a question of styling).

Almost all of the arrows are bidirectional, although is some cases the information flow is not bidirectional. In some cases the arrows are not consistent with the written algorithm.

The supplied pseudo code is unclear, very far from a proper one.
Big ratio of the code is not even code-like, instead closer to human procedure steps, without specified inputs, outputs, and other details.
Mixing the OOP like pseudo-code-like algorithm with procedural steps at such level is very strange.

In case of most functions return value is not defined, usual problem is the unclear storage of the results, although the style is OOP-like, so this should be much more precisely specified.
There are so many problems with the pseudo-code that it is considerable to remove them.
I will include some problems, but really it is very far from an acceptable one.

function AI_Driven_Smart_Event_API () is defined twice.

In function Model_Reasoning ()  on this high level the mentioned visualization of results is meaningless, even it is not clear how can the visualization be equal to reasoning.

In function Discover_Decisions () it is not clear what does the "Result" contain.

In function AI_Driven_Smart_Event_API ()

Why should a function be named "API"?

  decision_set=function Discover_Decisions ()
  if(decision_set==success)
     set flag=true

For a function call in pseudo code the "function" keyword should not be used.
It is not clear when a set of decisions means "success".

In function Event_Planning (event_guidelines)

do

   ....
    Apply_Permit (event_firm, participant_submission, venue_details)
    label = Event_Approval ()
  while (label==’TRUE’)
  Event_Approval ()
If the label equals 'TRUE' then the loop will continue, otherwise the loop ends and Event_Approval () will be called (if the label != 'True').
I guess this is not the intended behavior. Maybe you meant "until" instead of "while"?

Similar problem in function AI_Driven_Smart_Event_API ().

 

In function Event_Approval ()
    Site_Inspection (governance_firm, event_guidelines)
should be:
    inspect = Site_Inspection (governance_firm, event_guidelines)

    return label=TRUE
Here the literal TRUE is used instead of 'TRUE' which was used previously.

In function Smart_Application (input Event1)  :

This function does not return anything. The pseudo code is using an OOP approach. It creates local variables, which are not stored anywhere, eg. "e", "ds".

Event e= new Event () //Create instance event by the received event from Input
//Initialize the event and number of data sources
e= {Event1.communication, Event1.bulletin, Event1.advertisement}

The created Event instance "e" is overwritten instead of updating members of it.

In function Data_Transfer (Datasources ds[n])
    e.transfer (ds[n]) //Transfer the data using any of the given protocol and write the same consistently
    protocol={MQTT, HTTP, CoAP}

First does transfer with protocol which is set later?

The function seems to be incorrect.

In function Information_Extraction (Gateway g, Sensors s)

machine_learning_model(dataset) seems to be a function call, but model is usually an instance which has function members to call.

In function Mine_Knowledge ()

"choose finest model that gives better prediction" is not clear how, and whether this is chosen by the train results or test result or a validation set is used.

In function AI_Driven_Smart_Event_API () (2n variant)

Event_Approval ()
Event_Execute ()

The Execute call does not depend on the Approval?

In subsection 3.3 for some functions there is a structured description using bulleted list, for 3.3.5 the description is essay-like.
Subsection 3.4.1 has a not very clear structure, and even inconsistent or not clear numbers for percentages.

The description of the functions is very poor.

It is interesting that regression was mentioned while classification was not.
With machine learning background this is strange, it should be explained, why only regression, why not classification.

From the AI or ML perspective the academic level of the article is very low.

Author Response

Thank you for providing an opportunity to improve the manuscript. 

Response to the reviewer comments is provided in the attached file. Changes in the manuscript are highlighted (except for the Algorithm replacement) in the revised manuscript. 

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Thank you for the corrections.

There are still lot of problems with the article. Some of them were already mentioned previously in review. The authors are kindly asked to respect the efforts of the journal and reviewer, and provide quality work. It is also advisable to use a spell checker.

Some of the problems will be highlighted below.

 

Some parts of the article lack scientific attitude, objectivity and accuracy.

The proposed system and algorithm are not described and not validated appropriately.

 

In abstract

"Guided by our review, we shall  provide  a  framework using  machine  learning regression to effectively organise crowded events  during  the  ongoing  and  future crises."

organise -> organize

Why only regression is mentioned without classification?

Although there are algorithms which can be used both for regression and classification, in the field of machine learning we distinguish between the two purposes. If we use the ML to predict values, we call it regression, if we use the ML to choose category or decide between True and False, we call it classification. Usually e.g. using one_hot_encoder implies classification, also accuracy is used usually in classification, while in regression we use other metrics, like MSE, RMSE, MAE. In the human thinking classification can sometimes be fuzzy, which means can not only have very True or very False but any value between. It is also possible to fuzzify the binary class membership values, like in T. Tajti:Fuzzification of training data class membership binary values for neural network algorithms Annales Mathematicae et Informaticae. 52. pp. 217-228. DOI:10.33039/ami.2020.10.001 which means that the ML can "think" between False and True.

An other question is about choosing the model in your algorithm. It is correct that one algorithm can be better than an other but it can happen that some algorithms have very similar accuracy. In such cases it could be much more efficient to not only use the model with the momentary best accuracy but use multiple models with similar accuracy as a committee. Using these models to vote can achieve significantly better result, as it is show e.g. in T.Tajti:New voting functions for neural network algorithms Annales Mathematicae et Informaticae. pp. 229-242. DOI:10.33039/ami.2020.10.003 .

 

In section 2:

"Alpha,  Beta  Gamma,  and  Delta"

missing comma

 

The algorithm is still in bad format, it contains mistakes, and also fundamental problems.

 

In the function AI_Driven_Smart_Event

ds.CategoicalPart ---> ds.CategoricalPart

Mac_Accuracy ---> Max_Accuracy

Again, the Accuracy is usually used in classification, while in regression other metrics can be used. It is possible to define an Accuracy for regression, but in the article this is missing.

the line
For i=0 To Models.Count
seems to be semantically wrong because the loop has Models.Count+1 steps according to the usual meaning of this syntax

 

function Event_Execute is not a pseudo code

it does not use the other two functions

 

The source is not referenced for figures 1-4. Are they images made by the authors?

In 1.5

"We also provide validation of our algorithm."

I cannot find a proper validation in the article.

 

Figure 5. had a better image quality in V2, probably due to vector graphic format, while in V3 it is bad quality.

In subsection 3.3.5

"The accuracy of the models are computed and the  difference as DiffA and DiffB are computed by inverting the training and test splits."

What are DiffA and DiffB? Without defining them why is it important to name them?

 

Author Response

Response to Reviewer - file attached

Author Response File: Author Response.pdf

Round 4

Reviewer 2 Report

The overall quality of the revised document is much better now. Still there are some minor problems which could be corrected:

 

Figure 5. image quality is not very good. As mentioned previously, the image quality in V2 version was OK. Authors review states that it is corrected, but checking with various pdf viewers I still see the bad quality format.

In the algorithm I find that the authors try to keep the convention to close statements with semicolon, however it is missing in lines 277, 286.

 

Some lines of the algorithm are in italic, I cannot find a reason.

 

Line 265 and 364-365 are consistent, but 396-397 not consistent with them. test and validation are different purposes, validation is usually part of the training process. Line 441 has yet an other data split ratio (50%-50% instead of the previous 80%-20% and 80%-10%-10%). Why? This is not clarified.

 

line313: excecution - typo

 

line 43: Organisation - typo

line 166: fuzify - typo

line 301,305: event_Loc - is there any reason for the underscore?

 

sanitisation/sanitization - decide which form you use

 

Author Response

WE have addressed all comments of the Reviewer

Author Response File: Author Response.pdf