Submit to this Journal Review for this Journal Propose a Special Issue

Article Menu

Share Help Cite Discuss in SciProfiles

Open AccessArticle

Peer-Review Record

Harmonizing Heritage and Artificial Neural Networks: The Role of Sustainable Tourism in UNESCO World Heritage Sites

Sustainability 2023, 15(17), 13031; https://doi.org/10.3390/su151713031

by Alper Bozkurt

and Ferhat Şeker^*

Reviewer 1:

Peng Zhou

Reviewer 2:

Jian Wang

Reviewer 3: Anonymous

Reviewer 4: Anonymous

Reviewer 5:

Habib Benbouhenni

Sustainability 2023, 15(17), 13031; https://doi.org/10.3390/su151713031

Submission received: 1 June 2023 / Revised: 19 August 2023 / Accepted: 28 August 2023 / Published: 29 August 2023

Round 1

Reviewer 1 Report

sustainable tourism is an interesting and meaningful domain for AI application, MLP and RBF are applied and compared in this paper, however, some issues still should be clarified before publication.

Check the spelling, e.g. Shoa [16] in line 97, the name is shao

As topic of this paper a AI application, so please reduce the common introduction and focus on AI based analysis.

Both MLP and RBF are widely researched and applied in vary domain, so it is not necessary to introduce the them too much, just focus on the design and parameters , as well as the dataset preparing and processing that applied in paper.

Similarly，CAE and RAE are well known methods，please don't introduce too much about them. And as CAE and RAE are applied in this paper, please show the detail results of application in this research to prove the feature selection listed in table 2.

Please check the abbreviations, please don’t alternate between abbreviations and full terms. The full terms are used when firstly introduced, and the abbreviations are used in all last context. e.g. the RBF in the title of figure 2. and World Heritage Sites in line 22

As k-fold method is applied, are those k folds randomly selected, or how the authors keep Equivalence amount the k samples

Check the spelling, e.g. Shoa [16] in line 97, the name is shao

Author Response

Dear Reviewer,

We would like to express our sincere gratitude for your valuable feedback and constructive comments on our paper. Your insightful suggestions have greatly contributed to improving the quality of our research and we are grateful for the time and effort you have put into the reviewing process.

We have taken each of your comments and suggestions into careful consideration and made the necessary changes to the paper. To make the revisions clear, we have highlighted them in yellow colour for ease of identification.

Reviewer1: Check the spelling, e.g. Shoa [16] in line 97, the name is shao

Authors: We corrected the spelling of the name (see page 3, line 102)

Reviewer1: As topic of this paper a AI application, so please reduce the common introduction and focus on AI based analysis.

Authors: We removed some general information from the introduction.

Reviewer1: Both MLP and RBF are widely researched and applied in vary domain, so it is not necessary to introduce the them too much, just focus on the design and parameters , as well as the dataset preparing and processing that applied in paper. Similarly，CAE and RAE are well known methods，please don't introduce too much about them. And as CAE and RAE are applied in this paper, please show the detail results of application in this research to prove the feature selection listed in table 2.

Authors: The information about MLP, RBF, CAE and RAE has been shortened. The requested details have been added and highlighted (see page 10, lines 409-440).

Reviewer1: Please check the abbreviations, please don’t alternate between abbreviations and full terms. The full terms are used when firstly introduced, and the abbreviations are used in all last context. e.g. the RBF in the title of figure 2. and World Heritage Sites in line 22.

Authors: All of the terms were used when first introduced, and abbreviations were used and highlighted throughout the article in all subsequent contexts.

Reviewer1: As k-fold method is applied, are those k folds randomly selected, or how the authors keep Equivalence amount the k samples

Authors: Yes, the dataset in 10-fold cross-validation is divided into 10 folds randomly and we have added additional information and highlighted (see page 13, lines 476-480).

Once again, we appreciate your time and effort in providing useful feedback for our research. Thank you for your consideration and support.

Sincerely,

Authors

Reviewer 2 Report

It is difficult to find the theme and scientific significance. If comparing multilayer perceptron (MLP) and radial basis function (RBF) neural networks utilized in the classification of UNESCO World Heritage Sites, then how to do? By some experiment? what is dataset. And, the paper mentions a dataset, how to get it? what is its object and advantages?

Author Response

Reviewer2: It is difficult to find the theme and scientific significance. If comparing multilayer perceptron (MLP) and radial basis function (RBF) neural networks utilized in the classification of UNESCO World Heritage Sites, then how to do? By some experiment? what is dataset. And, the paper mentions a dataset, how to get it? what is its object and advantages?

Authors: The theme of the research is to explore the use of artificial neural networks, specifically multilayer perceptron (MLP) and radial basis function (RBF) neural networks, for classifying UNESCO World Heritage Sites and identifying key features for sustainable tourism.

The authors conduct experiments using a dataset of UNESCO World Heritage Sites. The dataset is a collection of information on UNESCO World Heritage Sites, such as name, inscribed date, longitude, latitude, region, and state (input variables) and the output parameters are cultural, natural, or mixed (See page 8).

The authors use feature selection methods to identify the most important variables for classification and compare the performance of MLP and RBF neural networks in terms of accuracy and evaluation time. The scientific significance of the study lies in its contribution to the international body of knowledge on sustainability and methodological approaches, as well as its potential usefulness to tourism researchers and practitioners interested in using AI to classify heritage sites for sustainable tourism.

Reviewer 3 Report

Please read the attachment. Thank you.

Comments for author File: Comments.pdf

The manuscript would benefit from thorough proofreading to address grammatical errors, improve sentence structure, and enhance clarity. Some sentences and phrases could be rephrased to ensure smoother readability.

Author Response

Dear Reviewer,

Reviewer3: ⎯ Introduction: + Please provide brief contributions in this section.

Authors: We provided brief contributions in introduction section and highlighted (see page 2, lines 78-83).

Reviewer3: + Please add a paragraph that introduces the outline of this manuscript.

Authors: We added the requested paragpragh and highlighted (see pages 2-3, lines 84-94).

Reviewer3: ⎯ Line 170: Please add Eq. (1) or Eqs. (1-2) to cite the equation in the text.

Authors: We added, please see page 6, line 259.

Reviewer3: ⎯ Eq. (1): please revise this equation. "r^2" should be "r2"

Authors: We revised, please see page 6, line 261.

Reviewer3: ⎯ Figure 1: the figure quality is poor. Please increase its resolution.

Authors: We increased the figure's resolution, please see page 7.

Reviewer3: ⎯ Eq. (3): Please replace "*" with "×"

Authors: We replaced it, please see page 9, line 352.

Reviewer3: ⎯ All equations should be cited in the text.

Authors: We cited them in the text, please see page 9; lines 351,357 and 377.

Reviewer3: ⎯ Eq. (5): Please replace "*" with "×"

Authors: We replaced it, please see page 9, line 378.

Reviewer3: ⎯ Table 2: the table and its title should be on the same page.

Authors: The table and its title are on the same page now (see page 11).

Reviewer3: ⎯ Figure 2: Please add the titles and units for both chart axes; the figure and title should be on the same page.

Authors: We added the titles and units for both chart axes (see page 12).

Reviewer3: ⎯ Figure 3: Please add titles and units for the charts.

Authors: We added the titles and units for the charts (see pages 14-15).

Reviewer3: ⎯ While the manuscript presents a trade-off between MLP and RBF in terms of evaluation time and accuracy, a more in-depth comparative analysis of the performance of these techniques would be valuable. Discussing the advantages and limitations of each method and providing a statistical measure.

Authors: We discussed them under the 3.materials and methods section and highlighted them (see pages 4-6, lines 170-244).

Reviewer3: ⎯ It would be helpful to provide more information about the dataset used in the study.

Authors: We provided detailed information about the dataset (see pages 10-11, lines 409-430).

Reviewer3: ⎯ Please check the accuracy column. Why did you choose it unchanged?

Authors: We explained it, please see page 13, lines 492-499.

Reviewer3: Literature review: Please add the theory and applications of OPTIMIZATION ALGORITHMS (lines 223-228) with more related work to intensely discuss the authors' mentioned method in this study. References they are limited. Please add more references to emphasize that your present study has made essential contributions to the field.

Authors: We added more studies related to optimization algorithms in the 2.literature review section and mentioned our study's contribution to the literature (see pages 3-4, lines 134-160).

Constructive Questions:

1. Methodology: How were the multilayer perceptron (MLP) and radial basis function (RBF) neural networks selected for this study? Were there any specific reasons or criteria for choosing these techniques over other classification methods? Additionally, could you provide more details on the experimental setup and parameters for training and evaluating the networks?

Answer: The selection of MLP and RBF networks for a particular study is typically based on several factors, including their suitability for the task at hand, their performance in similar studies or benchmark datasets, and the specific characteristics of the data being used. Researchers often consider the following aspects. MLP and RBF networks are versatile and can handle various classification tasks. Researchers may choose MLP if the problem involves complex relationships or nonlinear patterns, while RBF networks are often effective when localized patterns or interpolation is important. Researchers might refer to previous studies or established literature to assess the performance and suitability of MLP and RBF networks for similar classification tasks. Understanding the strengths and weaknesses of each technique helps in making an informed choice. In this study, Researchers consider their familiarity and expertise with MLP and RBF networks when selecting a suitable architecture. Prior experience and knowledge of these techniques can help ensure proper implementation and interpretation of results.

In this study, there are additional reasons for choosing MLP and RBF networks. These reasons could include their previous success in similar classification tasks, their compatibility with the available dataset and its characteristics, the specific research objectives and constraints, or the expertise of the researchers in utilizing these techniques. The authors provide a rationale for their selection, considering the unique context and requirements of their research.

The reason for using the dataset from the UNESCO official website is likely because it provides authoritative and reliable information about World Heritage Sites. The dataset from the official website ensures that the information used for classification is accurate and up-to-date, which is essential for conducting meaningful research in the domain of UNESCO World Heritage Sites. Based on the information provided, it seems that the choice of using the dataset from the UNESCO official website was primarily driven by the need for authoritative and reliable information about World Heritage Sites. The dataset obtained from the official website ensures that the information used for classification is accurate and up-to-date, which is crucial for conducting meaningful research in the domain of UNESCO World Heritage Sites. The decision to use the dataset from a reliable website ensures the quality and accuracy of the input data for the classification task.

2. Feature Selection: Apart from the region and date of the inscription, were there any other features considered for the classification of UNESCO World Heritage Sites? If so, what were they, and their respective contributions to the classification accuracy? Furthermore, what methodologies were employed by the Correlation Attribute Evaluator (CAE) and Relief Attribute Evaluator (RAE) to identify the region and date of inscription as the most prominent features?

Answer: No. The specific methodologies employed by the Correlation Attribute Evaluator (CAE) and Relief Attribute Evaluator (RAE) to identify the region and date of inscription as the most prominent features in the classification of UNESCO World Heritage Sites (WHS) are not provided in the given information. However, we provide a brief overview of these evaluators and their general methodologies. The Correlation Attribute Evaluator measures the correlation between each input attribute and the target variable (the classification labels) to determine their relevance or importance. It quantifies the linear relationship between the attributes and the target variable using measures such as Pearson correlation coefficient or mutual information. Higher correlation values indicate a stronger relationship with the target variable. The Relief Attribute Evaluator is a feature selection technique that assesses the relevance of each attribute based on its ability to discriminate between instances of different classes. It estimates the quality of an attribute by considering the differences in attribute values between nearest neighbours of the same and different classes. The algorithm calculates weights for each attribute based on the differences observed. Higher weights indicate greater relevance. Both CAE and RAE aim to identify attributes that contribute the most to the classification task. By measuring the correlation or discriminatory power of attributes, they provide insights into which features are most relevant for distinguishing between different classes of World Heritage Sites. The specific implementation details and methodologies utilized by CAE and RAE, including the choice of statistical measures and algorithms, are not mentioned in the provided information.

3. Discussion and Future Work: Considering the trade-off between evaluation time and accuracy observed between MLP and RBF, how might this impact the practical implementation of the classification model for real-world applications? Could you discuss potential scenarios where one technique might be more suitable than the other? Additionally, based on the findings of this study, what further research directions or improvements can be suggested to enhance the classification accuracy without compromising evaluation time?

Answer: The trade-off between evaluation time and accuracy observed between MLP and RBF networks has practical implications for the implementation of the classification model in real-world applications. There are some potential scenarios where one technique is more suitable than the other. If the classification model is intended for real-time or time-sensitive applications, such as online systems or applications that require immediate responses, RBF networks may be more suitable due to their faster evaluation time. The slightly lower accuracy of RBF networks can be acceptable in scenarios where speed is crucial. In applications where, high classification accuracy is of utmost importance, such as medical diagnosis or critical decision-making systems, MLP networks may be preferred despite their longer evaluation time. The higher accuracy achieved by MLP networks can outweigh the additional computational cost. In resource-constrained environments, such as embedded systems or devices with limited memory or processing power, RBF networks may be more practical due to their relatively lower computational requirements. They can provide a reasonable compromise between accuracy and resource usage. Based on the findings of the study, where MLP achieved higher accuracy but required more evaluation time compared to RBF, there are potential directions for further research and improvements to enhance classification accuracy without compromising evaluation time. First of all, Fine-tuning the architecture and hyperparameters of the MLP and RBF networks can lead to improvements in accuracy without significantly increasing evaluation time. Techniques such as regularization, parameter optimization, or architecture modifications can be explored to strike a better balance. Secondly, utilizing ensemble methods, such as combining multiple MLP or RBF networks, can potentially enhance the overall classification accuracy. Techniques like bagging or boosting can be employed to leverage the strengths of multiple models. Thirdly, this study identifies the region and date of inscription as prominent features for classification. Further exploration of additional relevant features, such as physical and environmental factors, could potentially improve classification accuracy while maintaining a reasonable evaluation time. Then, hybrid models that combine the strengths of MLP and RBF networks could be explored. For instance, using an RBF network as a pre-processing step to extract important features and feeding those features into an MLP network for classification might yield improved results. As a result, it is important to note that further research and experimentation are necessary to validate these suggestions and explore their effectiveness in improving classification accuracy without compromising evaluation time in the specific context of the study.

Once again, we appreciate your time and effort in providing useful feedback for our research. Thank you for your consideration and support.

Sincerely,

Authors

Reviewer 4 Report

The classification of UNESCO World Heritage Sites is an important tool for promoting sustainable tourism and ensuring the long-term conservation of cultural and natural heritage sites. Therefore, two commonly used techniques for classification problems, multilayer perceptron (MLP) and radial basis function (RBF) neural networks were utilized to define the pros and cons of their applications. Then, both Correlation Attribute Evaluator (CAE) and Relief Attribute Evaluator (RAE) identified the region and date of inscription as the most prominent features in the classification of UNESCO World Heritage Sites, according to the findings. As a result, a trade-off condition occurs for the classification between MLP and RBF in terms of evaluation time and accuracy. A large dataset for sustainable tourism can be classified with an accuracy higher than approximately 77% by MLP, higher processing time, and almost 76% via RBF with a much lower evaluation time.

- The authors worked on a topic, but the problem was not clearly defined.

- Although the ANN model is used to model the problem, theoretical book information about the MLP and RBF model has been given, and there is no clear information about modeling the problem.

- According to the results obtained, 77% performance was obtained for MLP and 76% for RBF, but these performance values cannot be accepted as very good results for today's values.

- The purpose of modeling the problem and the contribution of the result to science have not been given clearly and unequivocally.

- The available data were only used to make ANNs, there is no clear evidence why these 2 methods are used (although there are many different ANN models available).

For the above reasons, my opinion of this work is REJECT.

- The authors worked on a topic, but the problem was not clearly defined.

- Although the ANN model is used to model the problem, theoretical book information about the MLP and RBF model has been given, and there is no clear information about modeling the problem.

- According to the results obtained, 77% performance was obtained for MLP and 76% for RBF, but these performance values cannot be accepted as very good results for today's values.

- The purpose of modeling the problem and the contribution of the result to science have not been given clearly and unequivocally.

- The available data were only used to make ANNs, there is no clear evidence why these 2 methods are used (although there are many different ANN models available).

For the above reasons, my opinion of this work is REJECT.

Author Response

Dear Reviewer,

We would like to express our sincere gratitude for your valuable feedback. Your insightful suggestions have greatly contributed to improving the quality of our research and we are grateful for the time and effort you have put into the reviewing process.

Reviewer4: -The authors worked on a topic, but the problem was not clearly defined.

Authors: Optimization algorithms play a crucial role in the application of artificial neural networks (ANNs) for solving static optimization problems. While many ANN algorithms have been developed for static optimization, such as the Hopfield network (HN) and its derivatives, most of them do not involve a training procedure to adapt the weights of the networks. To overcome this problem, various optimization algorithms have been introduced in the design of neural networks, including particle swarm optimization and evolutionary algorithms. These algorithms have been proven to be feasible and effective in optimizing neural networks. In the context of UNESCO WHS, optimization algorithms can be used to analyze and optimize various aspects, such as the impact of sea-level rise on cultural heritage sites. Additionally, optimization algorithms can be applied to improve the accuracy of estimating tree diameter in historic gardens, which are UNESCO WHS. By incorporating optimization algorithms into the design and analysis of ANNs, researchers can enhance the performance and applicability of these networks in solving complex optimization problems. This paper presents a novel dataset for classifying UNESCO WHS and compares artificial neural networks in efficiency. While the input variables used in this study are similar to those used in previous research, some physical and environmental factors that were commonly employed in other studies are not included in the proposed dataset. The paper highlights the significance of selecting input variables carefully, considering research goals and available data. All in all, the proposed dataset and classification techniques represent a novel approach to UNESCO WHS classification that emphasizes the importance of carefully selecting input variables and balancing the trade-offs between accuracy and evaluation time. While the paper builds on previous research, it also introduces a new dataset and emphasizes the importance of taking into account specific research objectives as well as available data when selecting input variables for classification. Besides, the study identifies the most important variables in classification and compares the performances of artificial neural networks. As a result, the research fills gaps in the existing literature in many ways (see pages 3-4, lines 134-160).

Reviewer4: Although the ANN model is used to model the problem, theoretical book information about the MLP and RBF model has been given, and there is no clear information about modeling the problem.

Authors: We added more information about the model and discussed in-depth comparative analysis of the model's performance and gave some insight into why these techniques were chosen (see pages 4-6, lines 161-244).

Reviewer4: According to the results obtained, 77% performance was obtained for MLP and 76% for RBF, but these performance values cannot be accepted as very good results for today's values.

Authors: Whether the accuracy values of approximately 77% for MLP and 76% for RBF are acceptable or not depends on the specific context and requirements of the classification task for sustainable tourism. Generally, the acceptability of accuracy values can vary based on factors. The specific task and its implications play a crucial role in determining acceptable accuracy. Some applications may require higher accuracy levels for reliable decision-making, while others may have more lenient requirements. The domain in which the classification is being applied influences the acceptable accuracy. Some domains may inherently have more challenging classification problems, making achieving higher accuracy more difficult. It is important to consider the baseline or benchmark performance for the given classification task. Comparing the achieved accuracy to the baseline can provide insight into the relative performance and whether the obtained results are satisfactory. Accuracy should be considered in conjunction with other factors such as evaluation time, computational resources, and the cost of misclassifications. Balancing these trade-offs is crucial in determining the acceptability of accuracy values. In the specific context of sustainable tourism classification, without further information about the specific requirements and constraints, it is challenging to make a definitive judgment on the acceptability of the given accuracy values. However, it is worth noting that accuracy levels of approximately 76-77% can be considered reasonable depending on the factors mentioned above.

Reviewer4: The purpose of modeling the problem and the contribution of the result to science have not been given clearly and unequivocally.

Authors: While ANNs have the potential to be a useful instrument for the investigation of UNESCO WHS, their limitations in terms of data availability, reliability, and sensitivity to external factors must be carefully considered. Researchers should be cautious about relying solely on one ANN technique for analysis and should instead compare these models with other methods to gain a more comprehensive and nuanced understanding of these significant cultural and natural landmarks. The purpose of this investigation is, therefore, to compare multilayer perceptron (MLP) and radial basis function (RBF) neural network methods as a solution to this problem. Additionally, the paper aims to determine the most prominent features in the classification of world heritage sites through feature selection methods. This research significantly contributes to the international body of knowledge on sustainability and methodological approaches. The paper provides insights into the pros and cons of each method by comparing artificial neural network techniques for the classification of UNESCO WHS. This information may be useful to tourism researchers and practitioners interested in using AI to classify heritage sites for sustainable tourism. (see page 2, lines 69-83).

Reviewer4: The available data were only used to make ANNs, there is no clear evidence why these 2 methods are used (although there are many different ANN models available).

Authors: We discussed why these two methods were used (see pages 4-6, lines 170-244).

Once again, we appreciate your time and effort in providing useful feedback for our research. Thank you for your consideration and support.

Sincerely,

Authors

Reviewer 5 Report

1. In the abstract, what does UNESCO mean?

2. What is the contribution of the original paper?

3. What are the objectives of this work compared to some other works? new work?

4. Where is the difference between RBF and ANN (and MLP)?

5. What are the pros and cons of using RBF?

6. What about the learning algorithm?

7. The conclusion is too long, it should be rephrased.

Author Response

Dear Reviewer,

Reviewer5: 1. In the abstract, what does UNESCO mean?

Authors: We added the meaning of the UNESCO and highlighted it (see page 1, line 8).

Reviewer5: 2. What is the contribution of the original paper?

Authors: This research significantly contributes to the international body of knowledge on sustainability and methodological approaches. The paper provides insights into the pros and cons of each method by comparing artificial neural network techniques for the classification of UNESCO WHS. This information may be useful to tourism researchers and practitioners interested in using AI to classify heritage sites for sustainable tourism (see page 2, lines 78-83).

Reviewer5: 3. What are the objectives of this work compared to some other works? new work?

Authors: This paper presents a novel dataset for classifying UNESCO WHS and compares artificial neural networks in efficiency. While the input variables used in this study are similar to those used in previous research, some physical and environmental factors that were commonly employed in other studies are not included in the proposed dataset. The paper highlights the significance of selecting input variables carefully, considering research goals and available data. All in all, the proposed dataset and classification techniques represent a novel approach to UNESCO WHS classification that emphasizes the importance of carefully selecting input variables and balancing the trade-offs between accuracy and evaluation time. While the paper builds on previous research, it also introduces a new dataset and emphasizes the importance of taking into account specific research objectives as well as available data when selecting input variables for classification. Besides, the study identifies the most important variables in classification and compares the performances of artificial neural networks. As a result, the research fills gaps in the existing literature in many ways (see page 4, lines 147-160).

Reviewer5: 4. Where is the difference between RBF and ANN (and MLP)?

Authors: There is a comparison table for MLP & RBF (see Table 1 on page 8).

Reviewer5: 5. What are the pros and cons of using RBF?

Authors: We added the requested information about the techniques and highlighted it (see pages 4-5, lines 170-226).

Reviewer5: 6. What about the learning algorithm?

Authors: When it comes to learning algorithms for training neural networks like Multilayer Perceptron (MLP) and Radial Basis Function (RBF) networks, several options are available. Here are some commonly used learning algorithms for these network architectures. Backpropagation is the most widely used learning algorithm for training MLP networks. It utilizes gradient descent to update the network's weights based on the error between predicted and actual outputs. It iteratively adjusts the weights by propagating the error backward through the network, hence the name "backpropagation." Gradient descent is a general optimization algorithm that can be applied to various neural network architectures, including MLP and RBF networks. It updates the weights by iteratively descending along the negative gradient of the error surface. Variants of gradient descent include stochastic gradient descent (SGD), batch gradient descent, and mini-batch gradient descent. RBF Network Learning Algorithms: Training RBF networks typically involves two steps: clustering and parameter optimization. Clustering algorithms such as k-means or Gaussian mixture models are often used to determine the locations of the RBF centers. After clustering, the parameters, including the weights and widths of the RBF units, can be optimized using methods like least squares regression or iterative methods such as the Gauss-Newton algorithm. It is important to note that the selection of the learning algorithm depends on the specific network architecture, the problem being addressed, and the available data. Researchers may choose different learning algorithms based on their suitability, convergence properties, computational efficiency, and ability to handle specific challenges related to the task at hand.

Reviewer5: 7. The conclusion is too long, it should be rephrased.

Authors: Considering the recommendations from other reviewers, unfortunately, the conclusion part could not be shortened.

Once again, we appreciate your time and effort in providing useful feedback for our research. Thank you for your consideration and support.

Sincerely,

Authors

Round 2

Reviewer 1 Report

1）Check the equation and the variables, please use subscript to identify the variables, e.g. use C_j rather than Cj which is difficult for reading. In equation 2 line 265

2）Check all abbreviations, make sure when the full version of a term first appears in parenthetical text, place the abbreviation in square brackets after it. E.g RBFNN in lines 292

As MLP and RBF are two well research methods, the description in this version is verbose.

3）Please compare the results among the author’s methods and SOTA solutions if possible.

4）Authors try to explore the ANN application for promoting sustainable tourism and ensuring the long-term conservation, yet, in the paper, few results about that how ANN is applied to solve the classification problem of UNESCO WHS and and guide farther work , rather than the academic analysis of the two common methods.

Author Response

1）Check the equation and the variables, please use subscript to identify the variables, e.g. use C_j rather than Cj which is difficult for reading. In equation 2 line 265

Desired changes have been made.

2）Check all abbreviations, make sure when the full version of a term first appears in parenthetical text, place the abbreviation in square brackets after it. E.g RBFNN in lines 292

Desired changes have been made.

As MLP and RBF are two well research methods, the description in this version is verbose.

We understand your concern. Artificial neural networks (ANNs) have indeed been a relatively new addition to the field of tourism research. While methods like Multilayer Perceptron (MLP) and Radial Basis Function (RBF) have been extensively studied and used in various domains, their application and relevance in the context of tourism or sustainable tourism may not be widely known or understood by researchers in this specific field.

Given this context, it is important to provide a more detailed explanation of ANNs, MLP, and RBF in your research paper or study. This will help researchers in the tourism field who may not be familiar with these methodologies to understand their relevance and potential applications in their own research.

By providing a clear and concise description of ANNs, MLP, and RBF, you can bridge the knowledge gap and ensure that researchers in tourism or sustainable tourism can grasp the methodology and its implications more easily. This will help them make informed decisions about whether to adopt or further explore these methods in their own studies.

So, it would be beneficial to include an explanatory section or provide additional information to help readers understand the significance of ANNs, MLP, and RBF in the context of tourism research.

3）Please compare the results among the author’s methods and SOTA solutions if possible.

Thank you for your feedback and valuable suggestion. We truly appreciate your input and have carefully considered your request to compare the results among the author's methods and state-of-the-art (SOTA) solutions.

However, we would like to clarify that the classification of the UNESCO World Heritage Sites (WHS) in the context of sustainable tourism is a relatively new research area. As such, there are currently no other studies or SOTA solutions available for direct comparison in this specific domain.

Our research aims to address this gap by proposing a novel approach for the classification of UNESCO WHS in the context of sustainable tourism. By analyzing various criteria and factors, we have developed a comprehensive framework that allows for a more accurate and informed classification process.

While direct comparisons with other studies or SOTA solutions may not be possible at this stage, we believe that our research makes a significant contribution to the field by providing a pioneering framework for classifying UNESCO WHS in the context of sustainable tourism. Our findings offer valuable insights and practical implications for tourism management, conservation efforts, and policy-making.

The desired explanation has been added (See Page 2, Lines 74-88):

UNESCO WHS are invaluable cultural and natural treasures that require careful preservation and management for sustainable tourism. However, the vast number and diversity of these sites make their classification a complex task. Manual classification methods are time-consuming, subjective and inconsistent across different experts. Therefore, there is a need for more efficient, objective and accurate techniques to classify WHS sites based on their cultural, natural or mixed attributes. The purpose of this investigation is, therefore, to compare multilayer perceptron (MLP) and radial basis function (RBF) neural network methods as a solution to this problem. Additionally, the paper aims to determine the most prominent features in the classification of world heritage sites through feature selection methods. This research significantly contributes to the international body of knowledge on sustainability and methodological approaches. The paper provides insights into the pros and cons of each method by comparing artificial neural network techniques for the classification of UNESCO WHS. This information may be useful to tourism researchers and practitioners interested in using AI to classify heritage sites for sustainable tourism.

Reviewer 2 Report

This paper tries to solve some problems of sustainable tourism in UNESCO WHS by with artificial neural networks, and compares multilayer perception(MLP) and radial basis function(RBF) neural networks. But, you should give a clear definitions about the problem you want solve, at least including the conditions, and the objects. Thus, readers can understand your works.

Author Response

The desired explanation has been added (See Page 2, Lines 74-88):

Reviewer 3 Report

Please read the attachment. Thank you.

Comments for author File: Comments.pdf

Minor changes are needed.

Author Response

Comments on the Quality of English Language

Minor changes are needed.

Dear Reviewer,

Thank you for your feedback regarding the quality of the English language in our paper. We appreciate your attention to detail and have carefully reviewed the entire manuscript once again to address the minor issues raised.

During our thorough review, we made several minor changes to improve the clarity and cohesiveness of the paper. We have also taken the opportunity to paraphrase certain sentences to enhance readability and ensure that our intended meaning is accurately conveyed.

Reviewer 4 Report

Required revisions have not been completed in the manuscirpt. The gaps of the manuscript have also been given in the below:

The authors worked on a topic, but the problem was not clearly defined.

Although the ANN model is used to model the problem, theoretical book information about the MLP and RBF model has been given, and there is no clear information about modeling the problem.

According to the results obtained, 77% performance was obtained for MLP and 76% for RBF, but these performance values cannot be accepted as very good results for today's values.

The purpose of modeling the problem and the contribution of the result to science have not been given clearly and unequivocally.

The available data were only used to make ANNs, there is no clear evidence why these 2 methods are used (although there are many different ANN models available).

In this context, it cant be accepted. Because of the performance of the ANN is very low. Inputs and outputs have not been given in the manuscript. It has been a ANN model has been run and then results of the model has been given in the manuscript.

My decision is Reject. Sorry for this situation.

Required revisions have not been completed in the manuscirpt. The gaps of the manuscript have also been given in the below:

The authors worked on a topic, but the problem was not clearly defined.

Although the ANN model is used to model the problem, theoretical book information about the MLP and RBF model has been given, and there is no clear information about modeling the problem.

According to the results obtained, 77% performance was obtained for MLP and 76% for RBF, but these performance values cannot be accepted as very good results for today's values.

The purpose of modeling the problem and the contribution of the result to science have not been given clearly and unequivocally.

The available data were only used to make ANNs, there is no clear evidence why these 2 methods are used (although there are many different ANN models available).

My decision is Reject. Sorry for this situation.

Author Response

My decision is Reject. Sorry for this situation.

Dear Reviewer,

Thank you for reviewing our paper and providing your decision, even though it is a rejection. We appreciate the time and effort you have dedicated to evaluating our work.

Reviewer 5 Report

Thanks for the edits done, there are some notes that need to be amended. These notes are:

1. On line 96, you wrote "World Heritage Sites (WHS) are". Who writes best: Traditionally, WHS are. No need to repeat.

2. On line 134, you wrote “artificial neural net-

works (ANNs)". It is best to write "ANNs".

3. On lines 170 and 173, remove the "multilayer perceptron" and "Multi-172".

layer Perceptron" respectively, leaving the word MLP in their place.

4. On line 173, you have written "Radial Basis Function (RBF) networks". No need to repeat, you can adjust it with "RBF".

5. On line 246, please change the title to RBF technique.

6. On line 247, you have written "Radial Basis Function (RBF), which is". The same previous questions, there is no need to repeat it, please avoid it.

7. On line 299, write the address like this: MLP technique.

8. On line 300, you wrote “The Multilayer Perceptron (MLP) is a type of feedforward artificial neural network (ANN) widely used".

Please avoid boring repetitions. Same with line 409, 451, and line 453, 457, 487, 488, 529, 546, 550, 551, 580, and 593.

Check all these lines and the abbreviations are written the first time, and then only the keyword is used.

9. What about the properties of the neural networks used?

10. Please list the Training, Target, Mu,...of neural networks used.

thanks

Author Response

Thanks for the edits done, there are some notes that need to be amended. These notes are:

1. On line 96, you wrote "World Heritage Sites (WHS) are". Who writes best: Traditionally, WHS are. No need to repeat.