Forecasting the CO2 Emissions at the Global Level: A Multilayer Artificial Neural Network Modelling
Round 1
Reviewer 1 Report
This paper proposed a multilayer artificial neural network to forecast the CO2 emissions. This paper is well-written and the topic is an interesting. However, it should solve some problems before possible publication. Some issues must be handled are as follows:
Major comments:
- Contribution section should be rewritten. Contribution should include innovations, especially those that differ from other relevant research.
- In Introduction section, frame diagram of your paper should be provided for readers.
- How to determine ANN’s parameters, such as the number of layer (The grid search? You ran refer to DOI: 10.1016/j.measurement.2020.108468;) and lag orders of input. It is better to add more models (SVM or LSTM) and compare with the used ANN.
- Only MAPE is not enough, you can add MAE and RMSE to compare.
- Section 4：Section 4.1 was not found.
It will be useful to add MIV-based analysis to better understand your work. At least, you should discuss it briefly in the further research. Plz refer to the similar studies:
- https://doi.org/10.1016/j.energy.2021.120403.
- Induction Motor Multiclass Fault Diagnosis Based on Mean Impact Value and PSO-BPNN. SYMMETRY-BASEL
Minor comment:
- In Table 3, MAPE value during testing should be added “%”.
- The shortcoming and future directions should be added to the Conclusions.
Author Response
REVIEWER 1
Comments and Suggestions for Authors
This paper proposed a multilayer artificial neural network to forecast the CO2 emissions. This paper is well-written and the topic is an interesting. However, it should solve some problems before possible publication. Some issues must be handled are as follows:
Response: We thank the anonymous reviewer very much for all the constructive comments. We have taken every care to revise the manuscript following all the comments given by the reviewers. Here, we provide point-to-point response to the comments.
Major comments:
- Contribution section should be rewritten. Contribution should include innovations, especially those that differ from other relevant research.
Response: As suggested the contribution of the paper is added in the manuscript at page number 4, line no. 145-159 and page 5, line no. 181-197.
- In Introduction section, frame diagram of your paper should be provided for readers.
Response: Organization of the paper and the frame diagram are written in the revised manuscript.
- How to determine ANN’s parameters, such as the number of layer (The grid search? You ran refer to DOI: 10.1016/j.measurement.2020.108468;) and lag orders of input. It is better to add more models (SVM or LSTM) and compare with the used ANN.
Response: The aim of this paper is to evaluate the performance of MLANN for CO2 prediction. So the authors did not add other machine learning models. But it can be added in the future work.
- Only MAPE is not enough, you can add MAE and RMSE to compare
Response: Along with MAPE values, MAE and RMSE values are calculated and presented in Table 2 and 3.
- Section 4：Section 4.1 was not found.
It will be useful to add MIV-based analysis to better understand your work. At least, you should discuss it briefly in the further research. Plz refer to the similar studies:
- https://doi.org/10.1016/j.energy.2021.120403.
- Induction Motor Multiclass Fault Diagnosis Based on Mean Impact Value and PSO-BPNN. SYMMETRY-BASEL
Response: Section 4.1 is added. As the number of inputs is three in the current study, it is not necessary to reduce the features. Feature reduction is generally used when the dimension is very high.
Minor comment:
- In Table 3, MAPE value during testing should be added “%”.
Response: Correction is made as suggested.
- The shortcoming and future directions should be added to the Conclusions.
Response: Future directions and limitation of the proposed work are added in the conclusion section of the revised manuscript.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The authors employed ANNs for CO2 forecasting. The paper's subject is interesting. However, there are some issues to be improved to make the paper acceptable. My questions are raised in the following,
- The literature review is shallow. Please provide more papers regarding the paper's subject.
- The authors must compare the proposed approach with other forecasting methods, such as ARIMA models and ELM neural networks. Use as reference the structure of the paper https://www.mdpi.com/1996-1073/13/19/5190.
- The equations of performance measure MAPE must be presented in the paper.
- Once the adopted ANN has some hyperparameters for tunning, what is the method adopted for this purpose?
- Please improve the resolution of the figures. For me, it seems of lower quality.
Author Response
REVIEWER 2
Comments and Suggestions for Authors
Response: We thank the anonymous reviewer very much for all the constructive comments. We have taken every care to revise the manuscript following all the comments given by the reviewers. Here, we provide point-to-point response to the comments.
The authors employed ANNs for CO2 forecasting. The paper's subject is interesting. However, there are some issues to be improved to make the paper acceptable. My questions are raised in the following,
- The literature review is shallow. Please provide more papers regarding the paper's subject.
Response: As suggested, a detailed literature review is undertaken discussing the data used, methodology, and findings of the most recent papers. This literature review has guided the manuscript in findings the research gap. It appears at page number 2 - 4 of the revised manuscript.
- The authors must compare the proposed approach with other forecasting methods, such as ARIMA models and ELM neural networks. Use as reference the structure of the paper https://www.mdpi.com/1996-1073/13/19/5190.
Response: Currently our aim is to evaluate the performance of the MLANN model for CO2 emission forecasting. However, in our future work we will use autoregressive integrated moving average (ARIMA) and ELM models and will undertake a comparison of the two models.
- The equations of performance measure MAPE must be presented in the paper.
Response: Equation of MAPE is given in the revised manuscript.
- Once the adopted ANN has some hyperparameters for tunning, what is the method adopted for this purpose?
Response: There is no specific rule for tunning the hyperparameters. Trial and error method is used for tunning of the hyperparameters until the error value is minimized.
- Please improve the resolution of the figures. For me, it seems of lower quality.
Response: The figures are taken from the MATLAB software output screen directly. Better resolution figures are given in the revised manuscript.
Author Response File: Author Response.docx
Reviewer 3 Report
After having assessed the suitability for publication of the Manuscript ID: energies-1352917, having the title "Forecasting the CO2 emissions at the global level: A multilayer artificial neural network modelling", I have distinguished several elements that from my point of view should be made less confused and more comprehensible by the authors in view of improving the quality of the manuscript. Therefore, I have devised and wrote a series of comments to the authors of the manuscript under review.
In this paper, the authors forecast the CO2 emissions of the 17 emitting countries by making use of a multilayer artificial neural network model. The developed model uses Gross Domestic Product (GDP), urban population ratio, and trade openness, as predictors for CO2 emissions. By analyzing the obtained results, the authors remark which countries are expected to increase their emissions in near future, and which will continue to reduce their emissions.
The Manuscript ID: energies-1352917 is interesting and generally well-structured. However, the article under review will be improved if the authors address the following aspects in the text of the manuscript and reflect them clearly point-by-point within the coverletter:
- The main weak point of the manuscript. In my opinion, the main weak point of the Manuscript ID: energies-1352917 consists in the fact that even if the authors have cited a series of papers when presenting their obtained results in the "Results and discussion" section, these citations are used in order to sustain and justify the statements from the manuscript under review, but not in the purpose of devising a comparison between the authors' obtained results / devised approach and other existing ones from the literature. Therefore, this section of the manuscript does not reflect clearly the way in which the obtained approach can be perceived in perspective of previous studies that have tackled similar problems. This comparison is mandatory in order to highlight the clear contribution to the current state of knowledge that the authors have brought.
- Lines 27-106, the "Introduction" section – the state of knowledge. In the current form of the manuscript, the authors have performed a survey of what has been done up to this point in the scientific literature. However, in the "References" section, references 2 and 3 are identical, references 11 and 12 are identical and therefore, the total number of referenced papers is 18 instead of 20 (as appear in the current form of the paper). I consider that it will benefit the paper if the authors take into consideration more recent papers, in order to reflect the advancements in technology which took place over the last years, as in the actual form of the paper the cited papers range from 1993 to 2018, focusing more on the history and less on recent years. Actually, only 6 of the 18 referenced papers are published in the last five years, the rest of the papers being published before 2016. Moreover, from the 18 different papers contained by the "References" section in the actual version of the manuscript, 6 of them contain in the list of their authors at least one of the authors of the Manuscript ID: energies-1352917. I do not contradict the value of these papers, or their relevance in this context, but I consider that the article under review will benefit if the authors explain within the paper what was their citing criteria, based on which they have chosen the referenced papers. The authors should perform the literature review in a manner that emphasizes for each scientific work that was cited in the text of the manuscript the contribution that was made to the existing state of art, the approach that was employed along with a short description of the most important results that have been obtained along with the existing unsolved issues of the referenced studies. By doing so, the authors will be able to encompass their study in a broad body of knowledge and highlight unsolved problems that still exist regarding the subject that their manuscript addresses. The authors have cited some works from the current state of art, works that are relevant to their study, but the literature review conducted within the "Introduction" section should be extended as to cover a higher number of important scientific works from the literature and must be performed in a critical manner like I have explained in the above comments.
- The "Introduction" section – the gap in the current state of knowledge. After having extended and performed the literature survey in an appropriate manner, the authors will be able to pinpoint an exact deficiency, an unsolved problem that still exists in the current body of knowledge that their study addresses. This aspect will improve the manuscript under review on multiple plans, as the identified deficiency, the identified unsolved problem will offer great opportunities to highlight and prove, when discussing their results, the contribution, the advancement that the conducted research has brought to the existing state of knowledge. Afterwards, it will benefit to state the novel aspects of the conducted study.
- The "Introduction" section – presenting the structure of the paper. At the end of the Introduction section, the authors must present the structure of their paper under the form: "The rest of the paper is structured as follows: Section 2 contains…".
- The "Materials and Methods" section - presenting the devised approach. In order to help the readers better understand the methodology of the conducted research, in addition to the figures already presented, the authors should devise a flowchart when presenting their approach, a flowchart that depicts the steps that the authors have processed in developing their research and most important of all, the final target. This flowchart will facilitate the understanding and the reproducibility of the proposed approach and at the same time will make the article more interesting and help promote it to the readers.
- Issues regarding the multilayer artificial neural network approach. At Lines 230-240, the authors state: "It consists of two hidden layers with 9 and 3 neurons respectively." The authors should take into account a known fact from the scientific literature, that in the case of feedforward artificial neural networks, a single hidden layer containing an appropriate number of hidden neurons and an appropriate data division should have been sufficient in this case. The universal approximation theorem claims that the standard multilayer feed-forward networks with a single hidden layer that contains finite number of hidden neurons, and with arbitrary activation function are universal approximators. Since one hidden layer can provide a universal approximator, was it really necessary to use more than one hidden layer in the case of a feedforward artificial neural network?
- Issues regarding the neural network approach. As the authors have used a multilayer Artificial Neural Network approach, I consider that the authors must specify in the paper how often does the network need to be retrained/updated and how did they tackle the need of retraining/updating the network. How is the new data encountered stored for subsequent updates of the network? Meanwhile, the paper will benefit if the authors present more details regarding the results obtained during various tests, for all the different number of neurons and epochs tested and especially the training time for each test, until they have obtained the configuration that has provided the best results. The information can be summarized in a table and if it becomes too long, the authors can restrict it in the paper to ten main experimental runs, and a complete table with all the experimental runs can be inserted in the "Supplementary Materials" file of the article.
- Issues regarding the data division ratio. At Lines 230-231, the authors state: "Randomly selected 80% of data are used for training of the model and the rest 20% of data is used for testing of the developed model." The authors should explain in the paper the reasons for choosing this data division ratio. The paper will benefit if the authors present more details regarding the results obtained during various tests, for all the different tested ratio values, up to the moment when the chosen ratio has proven to be the best (or suitable) approach and what was the criterion/performance metric used in choosing this ratio.
- Issues regarding the equations within the manuscript. All the equations within the manuscript should be explained, demonstrated or cited, as a part of the equations have not been introduced in the literature for the first time by the authors and they are not cited.
- The "Simulation study" section – issues regarding the datasets. At Lines 218-219, the authors state: "The data for 9 countries under Group 1 and 8 countries under Group 2 are collected from 1960 to 2016". In order to be able to verify the correctness of the very important validations of the individual components that the authors state that they have performed against experimental data available in the literature, in the supplementary materials the authors must provide all the necessary details in order to allow other researchers to verify, reproduce, discuss and extend the obtained scientific results based on the obtained published results. Other researchers should not have to obtain and concatenate the data sets from various sources, risking in acquiring different datasets or datasets that have been normalized differently than the datasets on which the authors have performed their experimental tests and validations. The exact datasets that the authors have used when running the experimental tests and most importantly the validation process will be a valuable addition to the manuscript if they can be provided as supplementary materials to the manuscript as the authors must provide all the necessary details in order to allow other researchers to verify, reproduce, discuss and extend the obtained published results of the authors.
- The "Simulation study" section – details regarding the missing data or the abnormal values. The manuscript will benefit if the authors provide in the paper more details regarding the way in which they intend to solve the problems related to missing data or abnormal values if they are to occur.
- The generalization capability of the developed approach. Can the authors mention how much of their model is being influenced by the used data or to which extent the model can be easily applied to other situations, when the datasets are different? In this way, the authors could highlight more the generalization capability of their approach in order to be able to justify a wider contribution that has been brought to the current state of art.
- The advantages and disadvantages of the proposed approach. The authors should underline both the advantages and disadvantages of their proposed approach when compared with other valuable studies from the current state of art. When discussing their obtained results, the authors should emphasize not only the novel aspects and strong points of their developed method, but also to point out objectively the existing limitations of their method, possible circumstances that will hinder their method’s effectiveness and state clear and accurate directions they will pursue in their future research activities in order to extend the current research and overcome these limitations.
- Insights. The paper will benefit if the authors make a step further, beyond their approach and provide an insight when discussing their obtained results regarding what they consider to be, based on the obtained results, the most important benefits of the research conducted within the manuscript, taking also into account its practical applicability.
- The authors have submitted their paper to the Special Issue "Modeling Energy–Environment–Economy Interrelations" of the MDPI Journal Energies. As authors have submitted their paper to the Special Issue "Modeling Energy–Environment–Economy Interrelations" of the MDPI Journal Energies, I consider that the authors should strengthen the connection, relationship and main impact of their study on the energy domain. In the actual form of the paper, this connection is explicitly mentioned twice, namely at Lines 65-66: " The long-run relationship between CO2 emissions and its predictors such as GDP per capita, renewable energy consumption.", and at Lines 109-110: "The selection of countries in this study is based on the data compiled by the International Energy Agency". It will benefit the paper if the authors provide more details on this issue.
Minor remarks.
- The Figures and Tables within the manuscript. The authors must specify the measuring units (when applicable) along with the axes' titles within the figures or in the column names within the tables.
- Figure 1 is not referred in the paper.
- The content of the Supplementary Materials, Author Contributions, Funding, Institutional Review Board Statement, Informed Consent Statement, Data Availability Statement, Acknowledgments, Conflicts of Interest sections has not been filled in, being left in the form of the valuable indications from the Energies MDPI Journal's Template that should be applied effectively within the content of the manuscript, otherwise the readers are left only with good intentions without actions.
- The citations within the manuscript are marked using round brackets instead of square brackets, under the form "(Grossman &Krueger, 1995; Solarin, Al-Mulali, & Ozturk, 2017; Stern, 2004)" and therefore the citations are not in accordance with the recommendations of the Energies MDPI Journal's Template. According to this template, in the text of the manuscript, the reference numbers should be placed in square brackets [ ] and placed before the punctuation, for example [1], [1–3] or [1,3]. For embedded citations in the text with pagination, use both parentheses and brackets to indicate the reference number and page numbers; for example [5] (p. 10), or [6] (pp. 101–105).
- The "References" section. The authors should modify the references in accordance with the MDPI Energy Journal's Template. According to the Energies MDPI Journal's Template, the references must be numbered in the order of their appearance in the text (including citations in tables and legends) and listed individually at the end of the manuscript. In the actual form of the paper, the references are ordered in an alphabetical order instead of the recommended one. Please renumber and reorder the references in the "References" section, according to the recommendations.
- The acronyms within the paper. At Lines 16-17, the authors state: "The model uses GDP, urban population ratio, and trade openness, as predictors for CO2" Even if it is widely known in the scientific community, the GDP (Gross Domestic Product) acronym, as well as any other acronyms, should be explained the first time when they appear in the manuscript.
Author Response
REVIEWER 3
Comments and Suggestions for Authors
Response: We thank the anonymous reviewer very much for all the constructive comments. We have taken every care to revise the manuscript following all the comments given by the reviewers. Here, we provide point-to-point response to the comments.
After having assessed the suitability for publication of the Manuscript ID: energies-1352917, having the title "Forecasting the CO2 emissions at the global level: A multilayer artificial neural network modelling", I have distinguished several elements that from my point of view should be made less confused and more comprehensible by the authors in view of improving the quality of the manuscript. Therefore, I have devised and wrote a series of comments to the authors of the manuscript under review.
In this paper, the authors forecast the CO2 emissions of the 17 emitting countries by making use of a multilayer artificial neural network model. The developed model uses Gross Domestic Product (GDP), urban population ratio, and trade openness, as predictors for CO2 emissions. By analyzing the obtained results, the authors remark which countries are expected to increase their emissions in near future, and which will continue to reduce their emissions.
The Manuscript ID: energies-1352917 is interesting and generally well-structured. However, the article under review will be improved if the authors address the following aspects in the text of the manuscript and reflect them clearly point-by-point within the coverletter:
- The main weak point of the manuscript. In my opinion, the main weak point of the Manuscript ID: energies-1352917 consists in the fact that even if the authors have cited a series of papers when presenting their obtained results in the "Results and discussion" section, these citations are used in order to sustain and justify the statements from the manuscript under review, but not in the purpose of devising a comparison between the authors' obtained results / devised approach and other existing ones from the literature. Therefore, this section of the manuscript does not reflect clearly the way in which the obtained approach can be perceived in perspective of previous studies that have tackled similar problems. This comparison is mandatory in order to highlight the clear contribution to the current state of knowledge that the authors have brought.
Response: The ‘Results and discussion’ section has been revised to address this comment.
- Lines 27-106, the "Introduction" section – the state of knowledge. In the current form of the manuscript, the authors have performed a survey of what has been done up to this point in the scientific literature. However, in the "References" section, references 2 and 3 are identical, references 11 and 12 are identical and therefore, the total number of referenced papers is 18 instead of 20 (as appear in the current form of the paper). I consider that it will benefit the paper if the authors take into consideration more recent papers, in order to reflect the advancements in technology which took place over the last years, as in the actual form of the paper the cited papers range from 1993 to 2018, focusing more on the history and less on recent years. Actually, only 6 of the 18 referenced papers are published in the last five years, the rest of the papers being published before 2016. Moreover, from the 18 different papers contained by the "References" section in the actual version of the manuscript, 6 of them contain in the list of their authors at least one of the authors of the Manuscript ID: energies-1352917. I do not contradict the value of these papers, or their relevance in this context, but I consider that the article under review will benefit if the authors explain within the paper what was their citing criteria, based on which they have chosen the referenced papers. The authors should perform the literature review in a manner that emphasizes for each scientific work that was cited in the text of the manuscript the contribution that was made to the existing state of art, the approach that was employed along with a short description of the most important results that have been obtained along with the existing unsolved issues of the referenced studies. By doing so, the authors will be able to encompass their study in a broad body of knowledge and highlight unsolved problems that still exist regarding the subject that their manuscript addresses. The authors have cited some works from the current state of art, works that are relevant to their study, but the literature review conducted within the "Introduction" section should be extended as to cover a higher number of important scientific works from the literature and must be performed in a critical manner like I have explained in the above comments.
Response: Thank you for the comments. A detailed literature review is undertaken discussing the data used, methodology, and findings of the most recent papers. This literature review has guided the manuscript in findings the research gap. It now appears at page number 2 - 4 of the revised manuscript.
- The "Introduction" section – the gap in the current state of knowledge. After having extended and performed the literature survey in an appropriate manner, the authors will be able to pinpoint an exact deficiency, an unsolved problem that still exists in the current body of knowledge that their study addresses. This aspect will improve the manuscript under review on multiple plans, as the identified deficiency, the identified unsolved problem will offer great opportunities to highlight and prove, when discussing their results, the contribution, the advancement that the conducted research has brought to the existing state of knowledge. Afterwards, it will benefit to state the novel aspects of the conducted study.
Response: As suggested, we have followed the past literature on EKC to highlight the research gap in the revised manuscript. This is written in the Introduction section at page no. 4, line no. 145-159.
- The "Introduction" section – presenting the structure of the paper. At the end of the Introduction section, the authors must present the structure of their paper under the form: "The rest of the paper is structured as follows: Section 2 contains…".
Response: Organization of the paper is added in the revised manuscript at page number 5, line no. 194-197.
- The "Materials and Methods" section - presenting the devised approach. In order to help the readers better understand the methodology of the conducted research, in addition to the figures already presented, the authors should devise a flowchart when presenting their approach, a flowchart that depicts the steps that the authors have processed in developing their research and most important of all, the final target. This flowchart will facilitate the understanding and the reproducibility of the proposed approach and at the same time will make the article more interesting and help promote it to the readers.
Response: As suggested the Flow chart of methodology is added in the revised manuscript at page no. 8.
- Issues regarding the multilayer artificial neural network approach. At Lines 230-240, the authors state: "It consists of two hidden layers with 9 and 3 neurons respectively." The authors should take into account a known fact from the scientific literature, that in the case of feedforward artificial neural networks, a single hidden layer containing an appropriate number of hidden neurons and an appropriate data division should have been sufficient in this case. The universal approximation theorem claims that the standard multilayer feed-forward networks with a single hidden layer that contains finite number of hidden neurons, and with arbitrary activation function are universal approximators. Since one hidden layer can provide a universal approximator, was it really necessary to use more than one hidden layer in the case of a feedforward artificial neural network?
Response: It is a fact that ANN has generalization potentiality of replicating a nonlinear function. But the difficulty is in fixing number of artificial neurons in a layer as well as number of hidden layers. If a function is highly nonlinear in characteristics and by increasing number of layers and number of neurons one can achieve satisfactory convergence performance during training phase. However the performance degrades during testing phase. This over fitting problem needs to be resolved. Considering all these aspects, in the present case the desired two hidden layers with 9 and 3 neurons have been chosen by considering different combinations of no. of neurons and no. of hidden layers and observing the best possible testing performance.
- Issues regarding the neural network approach. As the authors have used a multilayer Artificial Neural Network approach, I consider that the authors must specify in the paper how often does the network need to be retrained/updated and how did they tackle the need of retraining/updating the network. How is the new data encountered stored for subsequent updates of the network? Meanwhile, the paper will benefit if the authors present more details regarding the results obtained during various tests, for all the different number of neurons and epochs tested and especially the training time for each test, until they have obtained the configuration that has provided the best results. The information can be summarized in a table and if it becomes too long, the authors can restrict it in the paper to ten main experimental runs, and a complete table with all the experimental runs can be inserted in the "Supplementary Materials" file of the article.
Response: It is true that retraining might be required for new data. In the present case two hidden layers with 9 and 3 neurons respectively have been chosen. However as suggested further simulation study is carried out by varying the ANN structure. The training and testing times as well as the performance achieved are obtained and listed in the following table
|
ANN structure |
Name of country |
Training time (in sec) |
Testing time (in sec) |
MSE in training |
MAPE value during testing(%) |
|
1hidden 5 neurons |
India |
6.138478 |
0.009161 |
2.3232e-04 |
4.0062 |
|
1hidden 6neurons |
6.216188 |
0.009189 |
3.5767e-04 |
5.3370 |
|
|
1hidden 7neurons |
6.123864 |
0.009349 |
2.9401e-04 |
4.2950 |
|
|
1hidden 8 neurons |
6.269911 |
0.008752 |
3.0132e-04 |
4.1285 |
|
|
1 hidden 9 neurons |
10.028355 |
0.010354 |
3.5327e-04 |
3.5209 |
|
|
2hidden 7 and 2neurons |
9.831674 |
0.009441 |
3.4736e-04 |
4.1261 |
|
|
2hidden 8 and 3 neurons |
10.636431 |
0.009528 |
3.4793e-04 |
3.8685 |
|
|
2hidden 9 an 3neurons |
9.585224 |
0.009823 |
3.1842e-04 |
2.9287 |
|
|
2hidden 9 and 4neurons |
9.871772 |
0.010052 |
2.0343e-04 |
3.1180 |
|
|
1hidden 5 neurons |
China
|
6.110742 |
0.009158 |
4.5298e-04 |
2.5947 |
|
1hidden 6neurons |
6.089258 |
0.009196 |
5.7419e-04 |
3.4048 |
|
|
1hidden 7neurons |
6.322200 |
0.009140 |
6.7688e-04 |
3.6029 |
|
|
1hidden 8 neurons |
6.152000 |
0.008976 |
5.0920e-04 |
3.1887 |
|
|
1 hidden 9 neurons |
6.229867 |
0.009011 |
4.8202e-04 |
3.7016 |
|
|
2hidden 7 and 2neurons |
9.389856 |
0.010075 |
4.3010e-04 |
1.9431 |
|
|
2hidden 8 and 3 neurons |
10.057381 |
0.009753 |
3.0071e-04 |
1.9622 |
|
|
2hidden 9 an 3neurons |
9.766929 |
0.009766 |
3.9001e-04 |
1.7896
|
|
|
2hidden 9 and 4neurons |
9.975999 |
0.009455 |
4.2449e-04 |
2.9416 |
|
|
1hidden 5 neurons |
Iran
|
6.913849 |
0.009312 |
5.7841e-04 |
4.6391 |
|
1hidden 6neurons |
6.887546 |
0.009127 |
7.1798e-04 |
4.3433 |
|
|
1hidden 7neurons |
7.074856 |
0.012755 |
7.9437e-04 |
4.3276 |
|
|
1hidden 8 neurons |
7.192900 |
0.009457 |
5.7418e-04 |
4.6925 |
|
|
1 hidden 9 neurons |
7.899810 |
0.012146 |
7.8194e-04 |
4.5488 |
|
|
2hidden 7 and 2neurons |
9.371471 |
0.010794 |
4.0064e-04 |
2.9805 |
|
|
2hidden 8 and 3 neurons |
9.362102 |
0.009566 |
6.9403e-04 |
2.9686 |
|
|
2hidden 9 an 3neurons |
9.417953 |
0.009692 |
4.1715e-04 |
2.3610
|
|
|
2hidden 9 and 4neurons |
9.448383 |
0.009704 |
6.3174e-04 |
3.4062 |
|
|
1hidden 5 neurons |
South Korea |
5.754868 |
0.008696 |
4.9800e-04 |
4.7633 |
|
1hidden 6neurons |
5.861521 |
0.008901 |
6.7324e-04 |
3.4051 |
|
|
1hidden 7neurons |
5.823516 |
0.008789 |
4.4774e-04 |
4.9455 |
|
|
1hidden 8 neurons |
5.971879 |
0.008950 |
3.6985e-04 |
4.5893 |
|
|
1 hidden 9 neurons |
6.232529 |
0.008714 |
4.4659e-04 |
5.4749 |
|
|
2hidden 7 and 2neurons |
7.720223 |
0.009725 |
4.7267e-04 |
4.0795 |
|
|
2hidden 8 and 3 neurons |
9.150815 |
0.009356 |
3.3289e-04 |
2.7722 |
|
|
2hidden 9 an 3neurons |
9.241007 |
0.009807 |
6.5473e-04 |
2.4803
|
|
|
2hidden 9 and 4neurons |
9.401891 |
0.010197 |
4.4097e-04 |
3.3424 |
|
|
1hidden 5 neurons |
Canada |
5.868935 |
0.008757 |
0.0011 |
3.8649 |
|
1hidden 6neurons |
5.918079 |
0.009310 |
0.0011 |
3.2082 |
|
|
1hidden 7neurons |
6.007617 |
0.009096 |
0.0011 |
3.9209 |
|
|
1hidden 8 neurons |
5.918384 |
0.008888 |
0.0011 |
3.9862 |
|
|
1 hidden 9 neurons |
5.953414 |
0.008799 |
0.0014 |
3.9108 |
|
|
2hidden 7 and 2neurons |
7.749690 |
0.009708 |
6.0841e-04 |
4.3843 |
|
|
2hidden 8 and 3 neurons |
7.810109 |
0.009941 |
0.0011
|
3.7945 |
|
|
2hidden 9 an 3neurons |
8.126898 |
0.009525 |
7.2345e-04 |
2.9358
|
|
|
2hidden 9 and 4neurons |
7.769944 |
0.009471 |
0.0012 |
3.9124 |
|
|
1hidden 5 neurons |
Indonesia |
5.935187 |
0.009459 |
0.0012 |
9.7618 |
|
1hidden 6neurons |
5.890987 |
0.008772 |
9.3132e-04 |
9.8918 |
|
|
1hidden 7neurons |
6.029074 |
0.009080 |
0.0019 |
10.5051 |
|
|
1hidden 8 neurons |
5.919254 |
0.011954 |
0.0023 |
10.3466 |
|
|
1 hidden 9 neurons |
5.993429 |
0.008863 |
5.1598e-04 |
10.8022 |
|
|
2hidden 7 and 2neurons |
11.321229 |
0.009985 |
0.0020 |
9.8449 |
|
|
2hidden 8 and 3 neurons |
8.231340 |
0.009959 |
0.0013 |
9.8369 |
|
|
2hidden 9 an 3neurons |
11.567121 |
0.015924 |
3.6109e-04 |
9.6898
|
|
|
2hidden 9 and 4neurons |
11.907859 |
0.009690 |
7.9649e-04 |
9.9591 |
|
|
1hidden 5 neurons |
Brazil
|
5.963350 |
0.009413 |
0.0012 |
6.4509 |
|
1hidden 6neurons |
5.937830 |
0.009012 |
0.0015 |
6.4964 |
|
|
1hidden 7neurons |
5.991229 |
0.009164 |
0.0012 |
6.0184 |
|
|
1hidden 8 neurons |
6.132805 |
0.009207 |
0.0011 |
6.2771 |
|
|
1 hidden 9 neurons |
6.088682 |
0.009066 |
0.0012 |
6.3698 |
|
|
2hidden 7 and 2neurons |
11.303435 |
0.009937 |
7.5015e-04 |
6.0302 |
|
|
2hidden 8 and 3 neurons |
9.845643 |
0.009931 |
0.0014 |
6.1265 |
|
|
2hidden 9 an 3neurons |
10.382500 |
0.010050 |
9.2622e-04
|
5.3345
|
|
|
2hidden 9 and 4neurons |
11.795995 |
0.01019 |
0.0012 |
6.1943 |
|
|
1hidden 5 neurons |
South Africa
|
5.850833 |
0.009198 |
0.0015 |
2.9419 |
|
1hidden 6neurons |
5.919399 |
0.009238 |
0.0012 |
2.8948 |
|
|
1hidden 7neurons |
5.884500 |
0.009034 |
0.0011 |
2.8201 |
|
|
1hidden 8 neurons |
6.110328 |
0.009050 |
0.0015 |
3.4969 |
|
|
1 hidden 9 neurons |
6.018903 |
0.008584 |
0.0017 |
3.5420 |
|
|
2hidden 7 and 2neurons |
8.343964 |
0.009915 |
6.9560e-04 |
2.9090 |
|
|
2hidden 8 and 3 neurons |
10.166757 |
0.009720 |
0.0011 |
2.8282 |
|
|
2hidden 9 an 3neurons |
9.733629 |
0.009733 |
6.1091e-04 |
2.7524
|
|
|
2hidden 9 and 4neurons |
9.618706 |
0.009444 |
9.5343e-04 |
2.7581 |
|
|
1hidden 5 neurons |
Mexico
|
5.858096 |
0.009000 |
0.0014 |
3.6678 |
|
1hidden 6neurons |
5.888978 |
0.009337 |
7.0457e-04 |
2.3329 |
|
|
1hidden 7neurons |
5.943035 |
0.008922 |
0.0011 |
2.5249 |
|
|
1hidden 8 neurons |
5.925326 |
0.009325 |
6.9130e-04 |
2.1518 |
|
|
1 hidden 9 neurons |
6.064403 |
0.008917 |
9.5070e-04 |
4.4129 |
|
|
2hidden 7 and 2neurons |
9.329162 |
0.009422 |
5.5520e-04 |
2.0929 |
|
|
2hidden 8 and 3 neurons |
9.155506 |
0.009456 |
4.5116e-04
|
2.0402 |
|
|
2hidden 9 an 3neurons |
9.758929 |
0.009660 |
1.6016e-04 |
1.9266 |
|
|
2hidden 9 and 4neurons |
9.605513 |
0.012369 |
8.4504e-04 |
2.5691 |
|
|
1hidden 5 neurons |
Turkey
|
6.054959 |
0.008867 |
2.4340e-04 |
2.9384 |
|
1hidden 6neurons |
6.178992 |
0.009523 |
3.0232e-04 |
2.3227 |
|
|
1hidden 7neurons |
5.896868 |
0.008963 |
1.1582e-04 |
2.3548 |
|
|
1hidden 8 neurons |
5.935108 |
0.009247 |
2.7431e-04 |
3.3400 |
|
|
1 hidden 9 neurons |
6.066213 |
0.009055 |
3.7015e-04 |
2.3665 |
|
|
2hidden 7 and 2neurons |
9.323097 |
0.009191 |
2.7164e-04 |
2.2197 |
|
|
2hidden 8 and 3 neurons |
9.258260 |
0.009341 |
1.5396e-04 |
2.8701 |
|
|
2hidden 9 an 3neurons |
9.335503 |
0.009375 |
1.5836e-04 |
2.1538 |
|
|
2hidden 9 and 4neurons |
9.484636 |
0.009525 |
1.7254e-04 |
2.8923 |
|
|
1hidden 5 neurons |
Australia
|
5.912477 |
0.009287 |
4.0076e-04 |
3.8219 |
|
1hidden 6neurons |
5.898434 |
0.009094 |
6.5776e-04 |
3.8112 |
|
|
1hidden 7neurons |
6.059746 |
0.009049 |
5.9981e-04 |
3.6756 |
|
|
1hidden 8 neurons |
6.188681 |
0.009060 |
6.6862e-04 |
3.6976 |
|
|
1 hidden 9 neurons |
6.863421 |
0.009375 |
6.4721e-04 |
3.6512 |
|
|
2hidden 7 and 2neurons |
9.219789 |
0.009911 |
5.0460e-04 |
3.6453 |
|
|
2hidden 8 and 3 neurons |
9.232531 |
0.009762 |
4.5971e-04 |
3.7155 |
|
|
2hidden 9 an 3neurons |
9.363223 |
0.009914 |
5.5587e-04 |
3.4001 |
|
|
2hidden 9 and 4neurons |
9.319407 |
0.009774 |
5.7673e-04 |
3.4514 |
Then a ranking of ANN structure is made by considering all performance, training and testing time. Then the best three structures are identified. In the present case the first best structure is presented.
- Issues regarding the data division ratio. At Lines 230-231, the authors state: "Randomly selected 80% of data are used for training of the model and the rest 20% of data is used for testing of the developed model." The authors should explain in the paper the reasons for choosing this data division ratio. The paper will benefit if the authors present more details regarding the results obtained during various tests, for all the different tested ratio values, up to the moment when the chosen ratio has proven to be the best (or suitable) approach and what was the criterion/performance metric used in choosing this ratio.
Response: The simulation is carried out with different data division ratios and it is observed that the 80-20% ratio is suitable for the proposed study as it is giving minimum MAPE value in all cases.
Group-I countries
|
Sl. No. |
Name of country |
Ratio of data division |
MAPE(in %) |
RMSE |
MAE |
|
1 |
India
|
80-20% |
2.9287 |
0.0198 |
0.0235 |
|
70-30% |
3.5062 |
0.0253 |
0.0214 |
||
|
60-40% |
5.0285 |
0.0345 |
0.0284 |
||
|
2 |
China
|
80-20% |
1.7896 |
0.0113 |
0.0150 |
|
70-30% |
5.3518 |
0.0350 |
0.0288 |
||
|
60-40% |
4.9046 |
0.0310 |
0.0263 |
||
|
3 |
Iran
|
80-20% |
2.3610 |
0.0262 |
0.0277 |
|
70-30% |
5.2414 |
0.0464 |
0.0399 |
||
|
60-40% |
4.5314 |
0.0359 |
0.0302 |
||
|
4 |
South korea
|
80-20% |
2.4803 |
0.0244 |
0.0324 |
|
70-30% |
4.0628 |
0.0422 |
0.0330 |
||
|
60-40% |
4.3806 |
0.0400 |
0.0330 |
||
|
5 |
Canada
|
80-20% |
2.9358 |
0.0244 |
0.0277 |
|
70-30% |
3.5630 |
0.0433 |
0.0344 |
||
|
60-40% |
4.0554 |
0.0423 |
0.0366 |
||
|
6 |
Indonesia
|
80-20% |
9.6898 |
0.0767 |
0.1077 |
|
70-30% |
7.4279 |
0.0852 |
0.0598 |
||
|
60-40% |
6.9903 |
0.0838 |
0.0463 |
||
|
7 |
USA
|
80-20% |
2.7168 |
0.0265 |
0.0308 |
|
70-30% |
5.3612 |
0.0574 |
0.0490 |
||
|
60-40% |
5.3558 |
0.0537 |
0.0507 |
||
|
8 |
Japan
|
80-20% |
3.5206 |
0.0214 |
0.0264 |
|
70-30% |
2.5840 |
0.0298 |
0.0244 |
||
|
60-40% |
3.0422 |
0.0349 |
0.0293 |
||
|
9 |
Saudi Arab
|
80-20% |
5.9153 |
0.0462 |
0.0535 |
|
70-30% |
6.4128 |
0.0550 |
0.0587 |
||
|
60-40% |
10.1447 |
0.0688 |
0.0549 |
Group-II countries
|
Sl. No. |
Name of country |
Ratio of data division |
MAPE(in %) |
RMSE |
MAE |
|
1 |
Brazil
|
80-20% |
5.3345 |
0.0330 |
0.0412 |
|
70-30% |
5.3642 |
0.0494 |
0.0414 |
||
|
60-40% |
6.2348 |
0.0501 |
0.0422 |
||
|
2 |
South Africa
|
80-20% |
2.7524 |
0.0279 |
0.0379 |
|
70-30% |
5.6019 |
0.0541 |
0.0475 |
||
|
60-40% |
3.6512 |
0.0372 |
0.0302 |
||
|
3 |
Mexico |
80-20% |
1.9266 |
0.0200 |
0.0224 |
|
70-30% |
2.5660 |
0.0302 |
0.0232 |
||
|
60-40% |
2.6386 |
0.0265 |
0.0227 |
||
|
4 |
Turkey
|
80-20% |
2.1538 |
0.0162 |
0.0209 |
|
70-30% |
2.7996 |
0.0233 |
0.0205 |
||
|
60-40% |
2.2536 |
0.0217 |
0.0151 |
||
|
5 |
Australia
|
80-20% |
3.4001 |
0.0367 |
0.0417 |
|
70-30% |
3.1685 |
0.0337 |
0.0292 |
||
|
60-40% |
4.5450 |
0.0422 |
0.0387 |
||
|
6 |
UK
|
80-20% |
3.5419 |
0.0410 |
0.0502 |
|
70-30% |
6.4151 |
0.0534 |
0.0470 |
||
|
60-40% |
4.0331 |
0.0361 |
0.0304 |
||
|
7 |
Italy |
80-20% |
8.8015 |
0.0653 |
0.0769 |
|
70-30% |
10.4971 |
0.1020 |
0.0869 |
||
|
60-40% |
9.3649 |
0.0915 |
0.0820 |
||
|
8 |
France |
80-20% |
3.8158 |
0.0241 |
0.0333 |
|
70-30% |
5.1282 |
0.0425 |
0.0335 |
||
|
60-40% |
5.2287 |
0.0419 |
0.0354 |
- Issues regarding the equations within the manuscript. All the equations within the manuscript should be explained, demonstrated or cited, as a part of the equations have not been introduced in the literature for the first time by the authors and they are not cited.
Response: In the revised manuscript, all equations have been explained. In addition, whatever equations taken from other sources have been cited. The explanation on equations and references of equations presented have been added at appropriate places of the revised manuscript. However, the equations given in the paper are standard equations easily available in neural network text books.
- The "Simulation study" section – issues regarding the datasets. At Lines 218-219, the authors state: "The data for 9 countries under Group 1 and 8 countries under Group 2 are collected from 1960 to 2016". In order to be able to verify the correctness of the very important validations of the individual components that the authors state that they have performed against experimental data available in the literature, in the supplementary materials the authors must provide all the necessary details in order to allow other researchers to verify, reproduce, discuss and extend the obtained scientific results based on the obtained published results. Other researchers should not have to obtain and concatenate the data sets from various sources, risking in acquiring different datasets or datasets that have been normalized differently than the datasets on which the authors have performed their experimental tests and validations. The exact datasets that the authors have used when running the experimental tests and most importantly the validation process will be a valuable addition to the manuscript if they can be provided as supplementary materials to the manuscript as the authors must provide all the necessary details in order to allow other researchers to verify, reproduce, discuss and extend the obtained published results of the authors.
Response: The details of data set, normalization used and number of training and testing sets have been made. As suggested, these additional materials have been included in the revised text as supplementary materials.
- The "Simulation study" section – details regarding the missing data or the abnormal values. The manuscript will benefit if the authors provide in the paper more details regarding the way in which they intend to solve the problems related to missing data or abnormal values if they are to occur.
Response: In this study the missing values are not taken into consideration. The missing data can be handled by doing mean, mode or median imputation.
- The generalization capability of the developed approach. Can the authors mention how much of their model is being influenced by the used data or to which extent the model can be easily applied to other situations, when the datasets are different? In this way, the authors could highlight more the generalization capability of their approach in order to be able to justify a wider contribution that has been brought to the current state of art.
Response: The proposed model can also be used for other data sets of different applications. However, as the model learns from its past data, training of the model is to be done once using the new data set. Then it can be used for different applications,
- The advantages and disadvantages of the proposed approach. The authors should underline both the advantages and disadvantages of their proposed approach when compared with other valuable studies from the current state of art. When discussing their obtained results, the authors should emphasize not only the novel aspects and strong points of their developed method, but also to point out objectively the existing limitations of their method, possible circumstances that will hinder their method’s effectiveness and state clear and accurate directions they will pursue in their future research activities in order to extend the current research and overcome these limitations.
Response: The proposed model is working well in most of the cases except two cases. (i) In case of Group 1 countries the MAPE value is 9.68% for Indonesia and the same value is 8.80% for the country Italy in Group 2. Further study will be carried out to reduce the percentage of error in these cases using other ANN based models like radial basis function neural network (RBFNN), recurrent neural network (RNN), Support vector machine (SVM), Extreme learning machine (ELM) or LSTM etc.
- Insights. The paper will benefit if the authors make a step further, beyond their approach and provide an insight when discussing their obtained results regarding what they consider to be, based on the obtained results, the most important benefits of the research conducted within the manuscript, taking also into account its practical applicability.
Response: The important practical application is - as we are able to predict the CO2 emission value it is beneficial for making alert to those countries whose emission rate is high or going to be high in near future and help them to make some strategies for reducing it as it is responsible for global warming.
- The authors have submitted their paper to the Special Issue "Modeling Energy–Environment–Economy Interrelations" of the MDPI Journal Energies. As authors have submitted their paper to the Special Issue "Modeling Energy–Environment–Economy Interrelations" of the MDPI Journal Energies, I consider that the authors should strengthen the connection, relationship and main impact of their study on the energy domain. In the actual form of the paper, this connection is explicitly mentioned twice, namely at Lines 65-66: " The long-run relationship between CO2 emissions and its predictors such as GDP per capita, renewable energy consumption.", and at Lines 109-110: "The selection of countries in this study is based on the data compiled by the International Energy Agency". It will benefit the paper if the authors provide more details on this issue.
Response: Thank you for this suggestion. In the revised literature review in the manuscript, the connection between CO2 emissions and renewable energy consumption is quite explicitly discussed.
Minor remarks.
- The Figures and Tables within the manuscript. The authors must specify the measuring units (when applicable) along with the axes' titles within the figures or in the column names within the tables.
Response: The suggestion is incorporated in the revised manuscript.
- Figure 1 is not referred in the paper.
Response : Fig. 1 is referred at the appropriate place in the revised manuscript.
- The content of the Supplementary Materials, Author Contributions, Funding, Institutional Review Board Statement, Informed Consent Statement, Data Availability Statement, Acknowledgments, Conflicts of Interest sections has not been filled in, being left in the form of the valuable indications from the Energies MDPI Journal's Template that should be applied effectively within the content of the manuscript, otherwise the readers are left only with good intentions without actions.
Response: All these sections are now suitably filled.
- The citations within the manuscript are marked using round brackets instead of square brackets, under the form "(Grossman &Krueger, 1995; Solarin, Al-Mulali, & Ozturk, 2017; Stern, 2004)" and therefore the citations are not in accordance with the recommendations of the Energies MDPI Journal's Template. According to this template, in the text of the manuscript, the reference numbers should be placed in square brackets [ ] and placed before the punctuation, for example [1], [1–3] or [1,3]. For embedded citations in the text with pagination, use both parentheses and brackets to indicate the reference number and page numbers; for example [5] (p. 10), or [6] (pp. 101–105).
Response: The reference section is revised as per the MDPI format.
- The "References" section. The authors should modify the references in accordance with the MDPI Energy Journal's Template. According to the Energies MDPI Journal's Template, the references must be numbered in the order of their appearance in the text (including citations in tables and legends) and listed individually at the end of the manuscript. In the actual form of the paper, the references are ordered in an alphabetical order instead of the recommended one. Please renumber and reorder the references in the "References" section, according to the recommendations.
Response: The reference section is revised as per the MDPI format.
- The acronyms within the paper. At Lines 16-17, the authors state: "The model uses GDP, urban population ratio, and trade openness, as predictors for CO2" Even if it is widely known in the scientific community, the GDP (Gross Domestic Product) acronym, as well as any other acronyms, should be explained the first time when they appear in the manuscript.
Response: This issue has been addressed.
Author Response File: Author Response.docx
Reviewer 4 Report
Remarks:
- In my opinion, the article does not add much new knowledge. The literature is too poor (only 20 references). It is mostly "old" literature, only 2 articles from 2018 and 1 from 2019.
- Change the citation method. The accepted way to cite literature in scientific journals in MDPI is to include the number from the References index in the text, rather than the authors' names.
- Correct formatting and errors everywhere, it is currently very careless:
CO2 –> CO2
SO2 –> SO2
NO2 –> NO2
L 30: 1.5 celsius -> 1.5 Celsius
Table 2. , L 264: Saudi Arab -> Saudi Arabia
Figure 5. (i): Saud Arab -> Saudi Arabia
Author Response
REVIEWER 4
Comments and Suggestions for Authors
Response: We thank the anonymous reviewer very much for all the constructive comments. We have taken every care to revise the manuscript following all the comments given by the reviewers. Here, we provide point-to-point response to the comments.
Remarks:
- In my opinion, the article does not add much new knowledge. The literature is too poor (only 20 references). It is mostly "old" literature, only 2 articles from 2018 and 1 from 2019.
Response: The manuscript is thoroughly revised. A detailed literature review is undertaken discussing the data used, methodology, and findings of the most recent papers. This literature review has guided the manuscript in findings the research gap. It appears at page number 2 - 4 of the revised manuscript.
- Change the citation method. The accepted way to cite literature in scientific journals in MDPI is to include the number from the References index in the text, rather than the authors' names.
Response: The MDPI referencing style is used in the revised manuscript.
- Correct formatting and errors everywhere, it is currently very careless:
CO2 –> CO2
SO2 –> SO2
NO2 –> NO2
L 30: 1.5 celsius -> 1.5 Celsius
Table 2. , L 264: Saudi Arab -> Saudi Arabia
Figure 5. (i): Saud Arab -> Saudi Arabia
Response: Language editing has been done and all these errors are corrected in the revised manuscript.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Some of my comments were not handled satisfactorily. The author did not take the comments seriously. Especially some important comments on the used ANN (parameters, performance verification, etc )in this study were not handled. Only easy problems were handled. So many technology details are not provided. The contribution is limited.
Generally speaking, the improvement is minor. Moreover, this manuscript (pdf file) is full of all kinds of marks. Maybe it is not ready for reviewers.
Author Response
REVIEWER 1
Comments of Reviewer 1
Some of my comments were not handled satisfactorily. The author did not take the comments seriously. Especially some important comments on the used ANN (parameters, performance verification, etc )in this study were not handled. Only easy problems were handled. So many technology details are not provided. The contribution is limited.
Generally speaking, the improvement is minor. Moreover, this manuscript (pdf file) is full of all kinds of marks. Maybe it is not ready for reviewers.
Response: We are extremely sorry for disappointing the reviewer. However, we tried our best to comply all the comments carefully given by the reviewers. Particularly, the two comments which were not fully addressed in the first round of revision have now been addressed and revisions are made. They are as follows.
We refer to the previous comment by the reviewer here –
- How to determine ANN’s parameters, such as the number of layer (The grid search? You ran refer to DOI: 10.1016/j.measurement.2020.108468;) and lag orders of input. It is better to add more models (SVM or LSTM) and compare with the used ANN.
The simulation is carried out by taking one and two hidden layers with different combination of neurons and the results are displayed in Table 1 in the Supplementary Materials. Accordingly, the text is revised in the manuscript in Para 1 and 2, on Page no. 16 in the revised manuscript.
We have referred to the paper Wu et al. (2021), DOI: 10.1016/j.measurement.2020.108468 and did the parameter tuning of the proposed ANN model. From Table 1 in Supplementary Materials, it is exhibited that comparing the Training time, testing time, MSE in training, and MAPE in testing, the proposed structure of MLANN model is better in comparison to other combinations of hidden layer and neurons. Similarly, the simulation is also carried out with different data division ratios and it is observed from Tables 2 and 3 in Supplementary Materials, that the 80-20% ratio is suitable for the proposed study as it gives the minimum MAPE value in all cases.
As suggested in Wu et al. (2021), other machine learning methods such as the SVM model is simulated and the resultant MAPE values are given in the Table below. It is observed that the MAPE values of all countries of Group-I and Group-II are not better in comparison to the proposed MLANN model. We have included the following Table as Table 4 in the Supplementary Materials. We have not added the methods of SVM and a detailed comparison between MLANN and SVM in the main text since it will require substantial expansion of the manuscript.
MAPE values for Group-I and Group-II countries using SVM with different kernels
|
Name of country |
SVM-Linear |
SVM-RBF |
SVM-Polynomial |
|
India |
4.9006 |
57.1999 |
114.7085 |
|
China |
29.2176 |
62.5550 |
265.2074 |
|
Iran |
36.0347 |
53.4743 |
37.3229 |
|
South Korea |
9.1170 |
52.5174 |
21.9167 |
|
Canada |
16.5645 |
17.8197 |
54.8594 |
|
Indonesia |
9.5794 |
54.5980 |
65.2664 |
|
USA |
14.7355 |
4.4472 |
37.3167 |
|
Saud Arab |
32.6173 |
55.2972 |
21.5335 |
|
Japan |
8.8881 |
5.3696 |
36.3962 |
|
Brazil |
5.5486 |
43.3320 |
8.2418 |
|
South Africa |
14.6854 |
39.4817 |
20.4860 |
|
Mexico |
5.6213 |
19.2069 |
33.2481 |
|
Turkey |
3.7920 |
47.7297 |
33.8005 |
|
Australia |
12.6662 |
33.7529 |
22.9515 |
|
UK |
11.9063 |
27.7873 |
241.0693 |
|
Italy |
19.1438 |
11.9933 |
21.6164 |
|
France |
10.1377 |
7.9741 |
35.7993 |
Further, referring to the earlier comment by the reviewer -
It will be useful to add MIV-based analysis to better understand your work. At least, you should discuss it briefly in the further research. Plz refer to the similar studies:
- https://doi.org/10.1016/j.energy.2021.120403.
- Induction Motor Multiclass Fault Diagnosis Based on Mean Impact Value and PSO-BPNN. SYMMETRY-BASEL
Response - We have referred to the above two papers in our manuscript. However, in our simulation we did not use lag orders of input. In our data set one tuple contains four columns out of which first three are the inputs to the model and the fourth value is the target or output value. In the same way we use all the tuples for training and testing.
Further, As the number of inputs is three in the current study, we have not tried to reduce the features. Feature reduction is generally used when the dimension is very high.
However, as suggested by the reviewer, we have discussed about this briefly in the further research at the end of the Conclusion section on Page no. 21.
According to the journal requirement, we have now uploaded a track-change version of the revised manuscript and a clean version of the manuscript in pdf.
Author Response File: Author Response.docx
Reviewer 3 Report
After having assessed the suitability for publication of the revised version of the Manuscript ID: energies-1352917, having the title "Forecasting the CO2 emissions at the global level: A multilayer artificial neural network modelling", I have noticed that when addressing a part of the issues signaled in my previous review report, the authors have added a few more tables (Table 4, Table 5 and Table 6) that spread along the pages 18-23.
As I have recommended to them in my first review report, I consider that, as the tables have become too long, in order to further improve the manuscript, the information summarized within the above-mentioned tables can be restricted in the manuscript to ten main lines, while the complete tables containing all data can be inserted in the "Supplementary Materials" file of the paper.
However, taking into account all the aspects that have been addressed in the revised version of the Manuscript ID: energies-1352917 and the point-by-point responses provided within the coverletter, I can conclude that the authors have addressed the most important signaled issues, therefore improving the manuscript in contrast to the prior submission.
Author Response
REVIEWER 3
Comments of Reviewer 3
After having assessed the suitability for publication of the revised version of the Manuscript ID: energies-1352917, having the title "Forecasting the CO2 emissions at the global level: A multilayer artificial neural network modelling", I have noticed that when addressing a part of the issues signaled in my previous review report, the authors have added a few more tables (Table 4, Table 5 and Table 6) that spread along the pages 18-23.
As I have recommended to them in my first review report, I consider that, as the tables have become too long, in order to further improve the manuscript, the information summarized within the above-mentioned tables can be restricted in the manuscript to ten main lines, while the complete tables containing all data can be inserted in the "Supplementary Materials" file of the paper.
However, taking into account all the aspects that have been addressed in the revised version of the Manuscript ID: energies-1352917 and the point-by-point responses provided within the coverletter, I can conclude that the authors have addressed the most important signaled issues, therefore improving the manuscript in contrast to the prior submission.
Response: We thank the reviewer very much for these constrictive ideas. We have revised the manuscript by summarizing the results contained in Tables 4, 5 and 6 in ten main lines. This revised text now appears in the 1st Para of Page no. 16. These tables have now been moved to Supplementary Materials.
Author Response File: Author Response.docx
Reviewer 4 Report
After reviewing the revisions made by the Authors, I currently have no further comments.
I conclude that the article may be published in the ‘Energies’.
Best regards!
Author Response
REVIEWER 4
We thank the reviewer very much for his/her constructive comments. They have helped us improve our manuscript.
Round 3
Reviewer 1 Report
No further comments. It can be accepted.
