Using Machine Learning Algorithms to Develop a Predictive Model for Computing the Maximum Deflection of Horizontally Curved Steel I-Beams
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors1- The models selected are good, but it would be better to compare one-bagging machine learning to these; maybe adding a random forest model will enhance the article.
2- More theoretical discussions on the selected machine learning models must be added to the article.
3- The main concern is the ratio of training to testing, which is 85% to 15%. In such cases, there is increased overfitting potential, which means that the metrics might not fully reflect the real performance of the models. For such cases, a 10-fold cross-validation is usually performed. However, I do not see any attempt to investigate this serious issue further. Consider reporting the 10-fold cross-validation results in your article along with the existing metrics.
4- There are many reference mistakes that make it difficult to follow. Revise them accordingly.
5- Normalize the results of Figures 22 and 27 for better interpretability.
6- Perform partial dependance plots to further understand how the machine learning interprets the data.
Author Response
Please find comments in the attached Word document.
The authors would like to thank the reviewer for his/her valuable comments and further commit to advancing the proposed research work that is found to be very promising in solving problems that could not be even considered in the past.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsEditorial:
Page 1 – Line 35; Page 2 – Line 77; Page 5 – line 198; Page 6 – lines 207-208; Page 7 – Lines 231 and 238; Page 8 – Line 257; Page 9 – Line 269, 281, and 290; Page 10 – Lines 298 and 309-310; Page 12 – line 352; Line Page 13 – Lines 358-360 and 378; Page 14 – Line 416; Page 15 – Line 429-430; Page 16 – Line 457; Page 18 – Lines 483, 497, and 499; Page 20 – Line 516; Page 21 – Lines 538, 546-547, 549, and 551; Page 23 – Line 576; Page 24 – Line 587: Correct the “Error! Reference source not found”
Technical:
1- What are the formulae used? This information should be provided for better review of the paper. This information is left out of the discussion.
2- Page 16 – Line 456 says “proposed predictive models”. What and where are these proposed models in the paper?
3- Page 16 – Lines 455-456 state “An analysis will be conducted to determine the performance of the various proposed predictive models.” Will be conducted? Then how were the analysis results produced and documented in this paper?
Comments on the Quality of English LanguageNone.
Author Response
Please find comments in the attached Word document.
The authors would like to thank the reviewer for his/her valuable comments and further commit to advancing the proposed research work that is found to be very promising in solving problems that could not be even considered in the past.
Author Response File: Author Response.docx
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors presented the prediction of performance of curved I-beam using various machine learning techniques. In general, this study is well written and well organized. Data design with both experimental and numerical investigation is reasonable. The methods and result discussion are well explained. Particularly, the validation section is plausible, which provides another perspective for various ML methods. The current format of the manuscript is acceptable.
If one thing the authors could be discussed should be real-world case application. Even though the authors tried to use experimental test to calibrate the FE simulation. The high uncertainty, for example, noise interference and other variances could significantly impact the ML prediction, particularly used for out-of-date prediction. In that case, it could be envisioned that the accuracy of various ML techniques could be shifted, or the DANN or Xgboot all may fail badly. This is generally issues for most data analytics for real-world applications.
Author Response
Please find comments in the attached Word document.
The authors would like to thank the reviewer for his/her valuable comments and further commit to advancing the proposed research work that is found to be very promising in solving problems that could not be even considered in the past.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors fairly addressed the comments.
Author Response
Once more, the authors would like to thank the reviewer for his/her valuable comments.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors’ responses are not persuasive, and they did not address the comments satisfactorily.
1- "Furthermore, the datasets that were developed for needs of this research work can be found through this link, whereas the proposed models can be directly developed by the reader." The link was not found in the revised paper.
2- How can this statement by the authors "As it was stated above, the authors avoided including very long formulae since it would be very difficult and painful for the reader to transfer the proposed predictive models to an Excel sheet...." be a response to the reviewer comment of: "What and where are these proposed models in the paper?". Long or short formulae must be presented to give credibility to the work; regardless if they are " very difficult and painful for the reader to transfer the proposed predictive models to an Excel sheet ".
3- "In order to avoid any misunderstandings, the sentence was changed to "An error analysis will be conducted..." " There is no misunderstanding. In the first draft, it was "An analysis will be conducted to determine the performance of the various proposed predictive models": How can the performance be submitted for publication (in the past) if "An analysis will be conducted to determine the performance...." (in the future). In addition, changing the sentence to "An error analysis will be conducted..." is not persuasive.
Author Response
Please find responses in the attached Word document. The authors would like to thank the reviewer for his/her valuable comments.
Author Response File: Author Response.docx
