Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Overall comment:

In the work, the authors developed an ANN model for South Korean demolition waste (DW) management, especially for old buildings. The authors used an artificial neural network algorithm with < 10 neurons in the hidden layer to derive individual input variables and optimal hyperparameters for each of the 10 DW types. All DWG prediction models achieved an average validation and test prediction performance (R²) of 0.970 and 0.952, respectively, with their ratios of percent deviation ≥ 2.5. Moreover, a Shapley additive explanations analysis revealed that DWG was most impacted by the floor area for all the DW types, with a positive correlation with DWG.

It is interesting to see ANN’s applications for waste generation management.

 

Comment 1: Why did the authors choose 7 features for the DW generation? How to justify this?

Comment 2: Does the year of the buildings play an important role for the DWG?

 

Comment 3: As the authors mentioned, how will the future ANN model evolve along with the change of building materials in the future?

Author Response

Thank you for your review of the completeness of this paper. The revision of this paper reflecting your review is as follows:

 

Point 1: Why did the authors choose 7 features for the DW generation? How to justify this?

 

Response 1: Thanks for your comment.

This study used 7 features to predict 10 demolition waste (DW) types. Various features are used to predict demolition waste generation (DWG). Existing studies also utilize various features to predict DWG, and the characteristics and number of features vary from study to study depending on the collected data. The features considered to predict DWG vary from study to study, but in general, gross floor area (GFA), structure, usage, and region are often considered important features that affect DWG (Please refer to 2.3.2. Input Variable Selection for Different Waste Types).

In addition to these features, this study included additional features (e.g., wall type, roof type, and number of floors) in the data. Because this study was to develop a DWG prediction model for each of 10 types of DWs, rather than a prediction model for the entire DWG generated in one building, additional features were needed, and 10 types of DWs were developed based on 7 features. Optimal feature sets were developed for each DW types. As shown in Table 2 and Table 5, this study confirms that various features were applied to each DW types based on 7 features.

While existing studies attempted to predict the entire DWG occurring in one building, this study developed independent prediction models and feature sets for 10 types of DWs that make up the entire DWG.

 

Point 2: Does the year of the buildings play an important role for the DWG?

 

Response 2: Thank you for your interest in DWG.

The age of a building is closely related to DWG, and there are many research documents related to this. Although this study did not separately mention literature related to age, age of buildins has a significant impact on DWG. This is because building materials used in buildings are influenced by technological advancements over time. For example, in the case of windows, aluminum or wooden window frames were widely used in Korea around 2000 or so. However, as technology advances, plastic windows with excellent insulation performance have been used in almost all buildings since 2000 year. These changes in building materials affect the composition and ratio of DW after the building is dismantled, and naturally also affect the DWG of DW types.

 

Point 3: As the authors mentioned, how will the future ANN model evolve along with the change of building materials in the future?

Response 3: I really appreciate your insights. The answer to your question is as follows, and we would like to inform you that this content was not reflected in the content because it was judged to be difficult to cover in the paper. We ask for your understanding regarding this matter.

Depending on the year the building was constructed, the composition and ratio of building materials will certainly continue to change. In this respect, future ANN models should consider the following changes.

1) Data characteristics: Future buildings have the potential for advancements in construction and material technology, as well as changes in environmental regulations and building rules. These changes may affect the composition and materials of the building, energy efficiency, durability, etc., so data characteristics must be taken into consideration. Therefore, future ANN models should be able to reflect new dismantled waste data.

2) Extraction of new input variables: New additional information that can reflect the environment or energy efficiency along with new material properties and construction methods can be considered.

3) ANN model architecture: The development of a new architecture based on data about future buildings is required.

4) Model verification and evaluation: Likewise, verification and evaluation of new models are required, and through this, the model must be adjusted or improved.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

 

This paper delves into the prediction of demolition waste (DW) in various building types using machine learning techniques, focusing on redevelopment areas in South Korea. It addresses an interesting topic with much room in the literature, while the authors employ state-of-the-art machine learning techniques and metrics. The paper is well-structured and well-written with significant contributions to the existing body of literature.

However, I have some comments for clarification and suggestions for improvements, which I believe would enhance the quality of the paper. Therefore, I propose a minor review of the paper so that the authors can address my questions or incorporate my comments.

Below, I provide detailed comments for the authors.

1)   I suggest using less definitive language about the exploitation of the paper’s results. For example, in the abstract, it is stated that ’’The study findings will enable demolition companies and local governments in making informed decisions for efficient DW management and resource allocation by accurately predicting the generation of various types of DW.’’ ‘’Will enable’’ could be replaced by other less definitive phrasing like ‘’could assist’’ or ‘’may assist’’.

2)     In the ‘’Introduction’’ section, it should be mentioned how the paper advances the existing body of literature from a methodological perspective. While the methodological steps are described in detail (something that the ‘’Materials and Methods’’ section already does), this aspect is overlooked in the introduction.

Currently, the description of the paper’s contribution in the introduction focuses on the characteristics of the examined region, with methodological contributions mentioned only in the "Discussion" section.

3)     In line 151, it is mentioned that ‘’to improve the prediction performance of AI models, a stable dataset must be constructed. The main purpose of building a stable dataset is to suppress the unwanted impact of distortions or outliers in the data.’’

However, the chosen normalization method doesn’t efficiently handle outliers in the dataset, as their effect is not eliminated. Typically, other techniques, including normalization methods (e.g., Z normalization), are utilized for this purpose. What was the rationale for selecting this normalization technique?

4)     In line 162, it is mentioned that ‘’ANNs are broadly classified into feedforward and feedback neural networks’’ However, the key differences between these two types of ANNs are not mentioned in the text. The basic definitions for each type should be provided.

5)     In Section 2.4, it is described that the analysis employs three evaluation metrics, namely MAE, RMSE, and R². First, what was the rationale for selecting these specific metrics? Also, if I understand correctly, the accuracy comparison between different models is based on the  R² metric, while the other metrics are used only for reporting purposes. This point needs clarification.

Furthermore, it is mentioned that ‘’ML model performance must be validated through a multi-criteria process to ensure that its accuracy is not exaggerated or distorted.’’ Does the study adopt such an approach? If so, the steps taken in that regard must be explicitly described. Otherwise, this part of the text must be removed to avoid confusion for readers.

 

Comments on the Quality of English Language

The paper is generally well-written with only a few minor issues.

Author Response

Thank you for your review of the completeness of this paper. The revision of this paper reflecting your review is as follows:

 

Point 1: I suggest using less definitive language about the exploitation of the paper’s results. For example, in the abstract, it is stated that ’’The study findings will enable demolition companies and local governments in making informed decisions for efficient DW management and resource allocation by accurately predicting the generation of various types of DW.’’ ‘’Will enable’’ could be replaced by other less definitive phrasing like ‘’could assist’’ or ‘’may assist’’.

 

Response 1: Thank you for your detailed advice. Sometimes when writing a paper, we do not pay careful attention to expressions. There may be various opinions on this point, but I think your opinion is appropriate. Therefore, this study reflected your opinion and revised the wording.

 

Point 2: In the ‘’Introduction’’ section, it should be mentioned how the paper advances the existing body of literature from a methodological perspective. While the methodological steps are described in detail (something that the ‘’Materials and Methods’’ section already does), this aspect is overlooked in the introduction.

Currently, the description of the paper’s contribution in the introduction focuses on the characteristics of the examined region, with methodological contributions mentioned only in the "Discussion" section.

 

Response 2: Thanks for your comments. We have supplemented the introduction to reflect your opinion. Please check lines 86 to 102 and line 117.

 

Point 3: In line 151, it is mentioned that ‘’to improve the prediction performance of AI models, a stable dataset must be constructed. The main purpose of building a stable dataset is to suppress the unwanted impact of distortions or outliers in the data.’’

However, the chosen normalization method doesn’t efficiently handle outliers in the dataset, as their effect is not eliminated. Typically, other techniques, including normalization methods (e.g., Z normalization), are utilized for this purpose. What was the rationale for selecting this normalization technique?

Response 3: Thanks for your advice You are right. We discovered a mistake in this part during the paper writing process. This part was revised again. Please refer to lines 173-177.

 

Point 4: In line 162, it is mentioned that ‘’ANNs are broadly classified into feedforward and feedback neural networks’’ However, the key differences between these two types of ANNs are not mentioned in the text. The basic definitions for each type should be provided.

Response 4: Thank you for your thorough review. This study has been strengthened by reflecting your opinions. Please refer to Lines 185-189.

 

Point 5: In Section 2.4, it is described that the analysis employs three evaluation metrics, namely MAE, RMSE, and R². First, what was the rationale for selecting these specific metrics? Also, if I understand correctly, the accuracy comparison between different models is based on the R² metric, while the other metrics are used only for reporting purposes. This point needs clarification.

Furthermore, it is mentioned that ‘’ML model performance must be validated through a multi-criteria process to ensure that its accuracy is not exaggerated or distorted.’’ Does the study adopt such an approach? If so, the steps taken in that regard must be explicitly described. Otherwise, this part of the text must be removed to avoid confusion for readers.

 

 

Response 5: Thank you for your review. The answer to this part is as follows. We sometimes evaluate the performance of the model additionally through other indicators when the value of R2 is the same or similar in the performance evaluation of the model. In addition, RMSE/MAE similarly evaluates the error of the model, so if it is difficult to judge from one indicator, it is possible to evaluate which model is better by referring to additional indicators.

Recently, in addition to single indicators, the multicriteria approach is used to evaluate model performance. For this reason, it is one of the additional approaches to verify ML model performance, and in many studies, Multicriteria (e.g., RPD and OBJ function) is adopted along with single indicators such as R squared, MAE, and RMSE.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

General Comment:

The paper presents a valuable contribution to the field of demolition waste (DW) management in South Korea by developing an AI-based model to predict the generation of various DW types. The models' performance is impressive, and the study has practical implications for demolition companies and local governments.

However, there are specific areas where the manuscript can be improved to enhance its quality and completeness.

 

Comment #1 - Literature review:

A critical component missing from the manuscript is the Literature Review section. A comprehensive literature review is essential to provide context and background for the research. To strengthen the study, a thorough literature review should be conducted, incorporating more recent and relevant references. This will help position the research within the existing body of knowledge and highlight its novelty.

 

Comment #2. Discussion:

The Discussion section could be improved by providing a deeper analysis of the findings in the context of existing literature. Authors should discuss the implications of their findings, both in terms of theoretical knowledge and practical applications. Consider comparing the results to previous studies and elaborating on what sets your research apart.

 

Comment #3. Conclusion:

The Conclusion section should be extended to offer a more detailed explanation of the practical implications of the research. Authors should respond to fundamental questions, such as how the research bridges the gap between theory and practice, its potential economic and commercial impact, influence on public policy, contribution to the body of knowledge, and impact on society, including public attitudes and quality of life.

 

Comment #4. Readability & overall structure:

The overall structure of the manuscript is coherent and logical. However, make sure that the flow between sections is smooth, and consider including clear transition sentences to guide the reader through the paper.

 

Final Comment:

In conclusion, the paper offers a valuable contribution to the field of DW management. To enhance the manuscript, however, it is crucial to perform a thorough literature review, strengthen the discussion section, and expand the conclusion to provide a deeper analysis of the practical implications of the research. Addressing these aspects will elevate the overall quality and relevance of the article.

Comments on the Quality of English Language

The manuscript demonstrates a good command of the English language; however, minor editing is needed to enhance overall readability. To improve the manuscript's coherence and logical flow between sections, consider incorporating clear transition sentences that guide the reader through the paper.

Author Response

Thank you for your review of the completeness of this paper. The revision of this paper reflecting your review is as follows:

Point 1: Comment #1 - Literature review:

A critical component missing from the manuscript is the Literature Review section. A comprehensive literature review is essential to provide context and background for the research. To strengthen the study, a thorough literature review should be conducted, incorporating more recent and relevant references. This will help position the research within the existing body of knowledge and highlight its novelty.

 

 

Response 1: Thanks for your advice. This study mentions the necessity of this study in the introduction through a review of existing studies related to this research. It may also be a good idea to organize a separate review section for existing research literature. However, since this study includes an introductory part, it does not seem necessary to organize an additional literature review section. It seems that there may be differences of opinion among reviewers on this matter, but I would appreciate it if reviewers would take this into consideration.

 

Point 2: Comment #2. Discussion:

The Discussion section could be improved by providing a deeper analysis of the findings in the context of existing literature. Authors should discuss the implications of their findings, both in terms of theoretical knowledge and practical applications. Consider comparing the results to previous studies and elaborating on what sets your research apart.

 

 

Response 2: Thanks for your review. We have strengthened the discussion section to reflect your comments. Please refer to lines 418-423 of the revised paper.

 

Point 3: Comment #3. Conclusion:

The Conclusion section should be extended to offer a more detailed explanation of the practical implications of the research. Authors should respond to fundamental questions, such as how the research bridges the gap between theory and practice, its potential economic and commercial impact, influence on public policy, contribution to the body of knowledge, and impact on society, including public attitudes and quality of life.

 

Response 3: Thank you for your review of the paper for its completeness. We have strengthened the content in the conclusion to reflect your opinion. Please refer to lines 470-489 of the revised paper.

 

Point 4: Comment #4. Readability & overall structure:

The overall structure of the manuscript is coherent and logical. However, make sure that the flow between sections is smooth, and consider including clear transition sentences to guide the reader through the paper.

 

Response 4: Thanks for your advice. We have revised the paper to reflect your opinions in many parts of the paper and make the sentences smoother. Please refer to the revised paper.

 

Point 5: In conclusion, the paper offers a valuable contribution to the field of DW management. To enhance the manuscript, however, it is crucial to perform a thorough literature review, strengthen the discussion section, and expand the conclusion to provide a deeper analysis of the practical implications of the research. Addressing these aspects will elevate the overall quality and relevance of the article.

 

Response 5: Thank you for your thoughtful review. We have reviewed this paper as a whole, and have revised the paper to reflect your opinions by adding the latest literature, strengthening the discussion section, and expanding the conclusion. Please refer to the revised paper.

Author Response File: Author Response.pdf