Optimizing Police Facility Locations Based on Cluster Analysis and the Maximal Covering Location Problem

Round 1

Reviewer 1 Report

Dear Author, 

Please consult the PDF-marked up comments. 

 

 

Comments for author File: Comments.pdf

Author Response

Dear reviewer, first of all, we would like to express our appreciation for reviewing our manuscript and for the extremely valuable comments. We reviewed our article and made improvements based on your observations. In order to facilitate the process for you, we have made some comments (in black) below your observations (in bold). In addition, the changes in the manuscript are marked up using the “TrackChanges”.

Our answers to your questions are in the attached file.

 

Kind regards,

The authors

 

Author Response File: Author Response.docx

Reviewer 2 Report

First of all, I think this research is not bad, and it should be able to provide a basis for the planning of local police facilities. but I still have some questions for you.

1.Cluster analysis can do different analysis methods, the most common ones are hierarchical clustering & k-means clustering. Different methods have their own significance. Please explain why did you choose to use k-means in this paper?

2. Page13, line495-501. I want to know that you selected 5 facilities as the main points to cover the radius, but they are divided into 10 groups in the cluster analysis, please explain the reason?

3. Page15, line443-550. You told us that the actual point distance is longer than the average point distance calculated by MIS, but in real life, if we want to change difference location of the facility, it is necessary to break through many restrictions, so what is your point  choose which will be prioritized in the paper ?

4. Page17-18 Conclusion mentions that MIS is helpful for interpreting events and providing managers with decision-making. but there is no questions and suggestions with discussion. So please tell me what kind policy is better that control crime rates in your paper?

Author Response

Dear reviewer, first of all, we would like to express our appreciation for reviewing our manuscript and for the extremely valuable comments. We reviewed our article and made improvements based on your observations. In order to facilitate the process for you, we have made some comments (in black) below your observations (in bold). In addition, the changes in the manuscript are marked up using the “TrackChanges”.

Our answers to your questions are in the attached file.

Kind regards,

The authors

Author Response File: Author Response.docx

Reviewer 3 Report

This article presents a methodology to optimally locate police stations. It is based on combining two known methods: a k-means algorithm to derive candidate locations, and a MCLP integer programming model to evaluate the best p-median combination of candidate locations. 

The problem is relevant and interesting but the methodology is not original. It has already been published in DOI: 10.1007/978-3-030-46224-6_7, by three of the same authors of this article.

Comparing this submission to MDPI Applied System Innovation to that of the DOI above, one can see no technical progress has been made:

- The idea is the same.

- The MIS implementation is the same.

- Tables and results are the same (modulo a few typos), as well as some figures.

- The main difference lies in a lot more talk in the present submission to ASI, which doesn't really add much to what's already available.

Since the research is not original, I am recommending rejection without possibility to ressubmit. Not only that, I have mathematical concerns on the usefulness of the methodology, as I could see from table 3 that combining the k-means and MCLP methods does not seem to provide much in terms of plus-values as compared to utilizing each of them separately, considering an MCLP with each crime scene as candidate location. 

Author Response

 

Dear reviewer, first of all, we would like to express our appreciation for reviewing our manuscript. We reviewed our article and made improvements based on the reviewer’s observations.

What we can argue about your observations is that the article has been significantly extended from the version that was previously published (first presented at a conference, with the extended abstract being published as a book chapter). The objective of this version is to present more methodological details, the MIS functionalities (still in the improvement phase, considering the interaction with the Security Secretariat) and other results, as well as further discussions.

 

Kind regards,

The authors

Round 2

Reviewer 1 Report

Please see the comments through the PDF-marked up. 

Comments for author File: Comments.pdf

Author Response

Dear reviewer,

first of all, we would like to express our appreciation for reviewing again our manuscript and for the valuable comments. We reviewed our article and made improvements based on these new observations. In order to facilitate the process for you, we have made some comments (in black) below your observations (in bold). In addition, the changes in the manuscript are marked up using the “TrackChanges”.

 

Kind regards,

The authors

Author Response File: Author Response.docx

Reviewer 3 Report

I have no problem with publishing an extended version of a conference paper. In fact, many journal articles are born this way. My point is that, despite authors' claims, the extension is like 99% motivational rather than technical, so I don't see much advancement to the state-of-the-art. Furthermore, authors should have been honest from the beginning and present the article as it is: an extended version of DOI: 10.1007/978-3-030-46224-6_7. Anyway, if the article is to be published, it is mandatory to cite that reference and mention that this article is an extended version thereof.

As for technical comments, in my previous review I raised an issue which I can now elaborate on. The mathematical concerns on the usefulness of the methodology comes from the fact that combining the k-means and MCLP methods does not seem to provide much in terms of plus-values as compared to utilizing each of them separately, considering an MCLP with each crime scene as candidate location. This conclusion was drawn from table 3, as I now show:

From the case study, 5 facilities are to be allocated and the table summarizes results from 5 to 1500 candidate locations, derived by using k-means.

Row 1 (5 candidates) is essentially running the k-means algorithm and discarding the MCLP step. It exhibits a 3 km coverage for 92% of crime occurrences.

Rows 2 and 3 (10 and 15 candidates) are MCLP optimal locations for respectively nCr = 252 and 3003 possible combinations of 5 facilities. Because the k-means centroids for 10 and 15 clusters differ from those for 5 clusters, suboptimal solutions arise (decrease of 92% to 90% in coverage): a 5 cluster k-means minimizes distances to occurrences, so it is natural that more under 3 km ocurrences are covered than when different candidate locations are enforced.

Last and next-to-last rows (1200 and 1500 candidates) are very similar to running the MCLP with each crime occurrence as a candidate location, as authors themselves recognize in lines 589-590 (call this a complete MCLP). Since minimizing distance is not always the same as maximizing coverage, an improvement on row 1 is found: 92% to 96% coverage.

Given the above, it would seem that either a simple k-means or a complete MCLP could be used instead of the proposed methodology and combining the two is, strictly speaking, not necessary. It may be preferrable to its alternatives under specific circumstances, since mathematically neither the k-means nor the complete MCLP guarantee absolute optimal coverage (counterexamples are easy to construct). This would however require new research on sensitivity analysis, more (and larger) case studies, and extensive numerical testing. Perhaps this article should have focused on those aspects. That would have made it a much more interesting and valuable piece of research.

Other issues I see with this article are as follows:

1. The writing can be very confusing, even unreadable in some sections, and is often dominated by linguistic contamination. It does not meet the standards for a scientific paper and a thorough revision is needed. Section 2.1 is especially painful to read. Also, "Where I've been robbed" is probably a better translation of "Onde fui roubado" (unless you want to put it as a question, in which case "Where have I been robbed" is probably the thing you want).

2. The article has many bits of information which are repeated or really not necessary. That text space would be much better used explaining some methodological details (e.g., a workflow of the methodology). Here are some examples (many more exist): 

- Lengthy introduction. I think it's pretty obvious for everyone crime fighting is important.

- Lines 294-306.

- Lines 311-314.

- Lines 371-372.

- Lines 390-391. Just cite Tierney if you really want to mention him.

- Table 1.

- I am not totally familiar with MDPI Applied System Innovation style, but I think sections 4.1 and 4.2 which describe the MIS implementation could be removed or at least greatly reduced. 

- Table 2 has typos in the last line: 500 should be 5.00 and 1638 should be 1683, as can be seen from table 1 of DOI: 10.1007/978-3-030-46224-6_7.

3. Line 443. Four analyses? Does that mean the MIS can optimize the candidate locations in more than one way? Why was that not mentioned in the text?

4. Eqns. (1-5). Standard mathematical notation is not followed. Variables should be italicized, both as equations and in-text. Also, why is the desired service distance 'S' between apostrophes in-text?

5. Line 234 and eq. (1). The a_i of line 234 is probably the w_i of eq. (1).

6. Line 311. Of all the citations in this line, only [52] is really relevant, and even then only slightly. That reference uses k-means to start an optimization algorithm, [51] has little to do with this work, and other references are just preliminary work.

Author Response

Dear reviewer,

first of all, we would like to express our appreciation for reviewing our manuscript and for the extremely valuable comments. We reviewed our article and made improvements based on your observations. In order to facilitate the process for you, we have made some comments (in black) below your observations (in bold). In addition, the changes in the manuscript are marked up using the “TrackChanges”.

 

Kind regards,

The authors

Author Response File: Author Response.docx