Review Reports

Round 1

Reviewer 1 Report

I would like to congratulate your work which shows a high effort in the collection of data and its computing. However, some changes should be done:

• Text quality should be enhanced for the easiness of its reading: See the abundance of words in page 2-3-4 or from line 200-238 that makes the work a little hard and boring. Paragraph from line 54 to 59 is written with few punctuation marks and it´s not clear its understanding.
• Some statements should be referenced, otherwise is just the authors´ opinion which may considered biased: "Such strategic combinations can 70 reduce the total error of learning models (including bias and variance errors) or enhance 71 the performance of single-model classifiers" in line 70 or "Ensemble learning.....) from line 72 to 77. "However, it is certain...." in line 149.
• Bagging-C5 and Bagging-C50 may be read throughout the work.
• 3.1 Data source: It´s difficult to be understood why some data like GDP or CPI are needed to obtain results. 
• In the same line than former appreciattion. In Table 1 and "Customer broker" section: Why is needed some data like "number of days from the previous importation or rate of change of number of days taken for importation". Why are needed 125 factors in Table 1? It´s not very well explained.

Author Response

Our replies to reviewer 1 are listed in the cover letter file.

Author Response File: Author Response.pdf

Reviewer 2 Report

Title: “Ensemble learning–models for food safety risk prediction”

Manuscript number: SUSTAINABILITY_1434400

The manuscript present models for food safety risk prediction by employing artificial intelligence tools. Overall, the presented are sound and effectively the approach may be of great interest to those in charge of food safety.

Comments:

1. What are the limitations, if any, of the employed approach? Kindly discuss.

Author Response

Our reply to the comments of Reviewer 2 is listed in the the cover letter file.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors presented and described the models of risk prediction in order to verify their usefulness for border food inspection. To achieve this goal a confusion matrix to calculate predictive performance indicators was formed. The results have proved ensemble learning achieved more precise prediction data than any single algorithm. The usefulness and novelty of such an approach are unarguable. My main concern is about the potential implementation of these results – can the authors comment in more detail this issue in the corrected version of the manuscript.

Author Response

Our reply to Reviewer 3 is listed in the cover letter file.

Author Response File: Author Response.pdf