Influencing Factors and Prewarning of Unsafe Status of Construction Workers Based on BP Neural Network

Round 1

Reviewer 1 Report

Very interesting study on the factors that can influence the unsafe behavior of construction workers, a relevant and current topic. For the analysis of the data collected, the authors used the toolbox of neural networks provided by MATLAB, presenting results of a specific study group. However, some comments must be made:

• Keywords are too long. Four keywords made up of short phrases.

• The reported data of causalities in China in 2019, given the size of the country, seems very low (904). On the other hand, the comparison with Australia does not seem very adequate. They are different realities.

• The text lacks a major revision, for example: line 66, “Wu Fan et al. established…” instead of “Wu Fan et al. [7] established…”. The same goes for lines 70, 76, 80, etc. On line 154 the initials BP must be preceded by “BackPropagation” instead of backpropagation (BP) …

• The references presented must be reviewed as they are not complete. The lack of DOI or URL makes it impossible to validate and consult.

• The tables are, almost entirely, broken, that is, header on one page, body of the table on the next page (line 144 for example).

• Table 1 is very similar to Table 12 on page 17, line 413.

• The results shown in Table 4 require explanation. What values are used to calculate RI?

• Line 235 presents an acceptance criterion without justification for the statement (CR<0.10). Why?

• Figure 2 does not exist. It is not displayed (line 245).

• In Equation (6) the variables are not defined. The same goes for the expression on line 255.

• Figure 3 can be improved.

• In line 304 the authors refer that the results presented in Table 5 are the result of combining equation (2) and (4). Equation (2) is CI and (3) is RI. Equation (4) is the combination of (2) and (3) CR=CI/RI. The λmax, eigenvector, CI and CR data are not explicit as to their origin.

• Table 7 would be more understandable if lines separating the information were created. In the Worker information (Area) the sum of the % is less than 100%.

• The blocks shown in Figure 4 are a black box that does not allow viewing the mathematical approach used in the neural network. For an expert it may be enough. For others, it becomes incomprehensible, which leads to a misunderstanding of the values in Table 10 (Hidden Layers).

• The information presented in Table 15 is a repetition of the information presented in Table 14 presented in another format. The data in Table 14 are sufficient for understanding the factors by consulting Table 12 or Table 1.

• The results presented in Table 16 are fully comprehensible. It appears that those on the left are “Before” and those on the right are “After”. It will be.

• Conclusions are presented in a general way but supported by the four workers studies and the different points of view of experts and workers without focusing on a possible common focus.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

What authors mean by unsafe behavior of workers is not enough described. There is no relation with accidents or injuries at work. Due to this, this paper looks quite theoretical.

One foctor is missing: how workers learn from incidents and accidents. Using only their experience duration is not sufficient. If they experience few dangerous situations, there are less prepared than workers who had to cope with many hazardous situations and forged their own safety culture

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thank you for clarifying the comments made. The changes introduced by the authors made the paper more understandable. Even so, I must make a few comments.

1- The tables are still broken, very extensive.

2- Check the writing of the document as there are still some mistakes, for example, line 177 “judgment matrix of.” cut "of".

3- The equations must be renumbered. Line 272 the equation number becomes (8). The following should be updated. On line 378 the new equation will have to be (11).

4- Lines 451 and 459 correct “steelworks” to “steel workers”.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

There are important errors in the conclusion:

They all believe that the three factors of protection measures, group cohesion and security awareness are very insignificant (should be significant)

safety compliance and security culture are not insignificant (should be significant)

The sentence below should be removed and replaced by a sentence promoting the factors identified as the strongest in your study to be reinforced by management and companies' owners (training, observation of hazardous attitudes by peers, rewards, ...).

Considering some factors will change if accident occurs in the project . Therefore, it is suggested that some measures should be took to further ensure the safety of workers, such as strengthening safety protection, reducing safety hazards, and conducting safety education for workers through VR. Workers will also improve their safety awareness and learn more about safety protection.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf