Construction and Application of Carbon Emissions Estimation Model for China Based on Gradient Boosting Algorithm

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study is devoted to effectively building machine models for CO2 emission prediction at the county level, and quantifying the contributions of different driving factors. Here are some of my comments to enhance the research.

 

Abstract

The shortcomings of existing research and the key innovations of this study should be clearly pointed out.

 

Introduction

The introduction of GBDT, XGBoost, and LightGBM models is abrupt. Please provide the potential evidence that the gradient boosting algorithms are suitable for predicting CO2 emissions at county level.

 

Materials and Methods

Please add a description of the study area and the corresponding figure.

In this section, the authors should fully expound the innovative combination of this study, rather than simply describe the classic models.

The authors should describe in detail which data are input into the model in what form to get the prediction results.

Please explain the reasons why the ARIMA model was used.

Please add a detail description about the verification of model results, including 2008-2017 and 2018-2027, respectively.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors This study investigated a carbon emission estimation model based on three gradient boosting algorithms: Gradient Boosting Decision Tree (GBDT), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM). This method is robust for the carbon emission estimation model construction and application. It is interesting and useful to the field of carbon emission estimation and geospatial analysis. However, there are some issues or questions should be addressed: (1) More advanced feature selection methods, such as recursive feature elimination or feature importance-based selection using tree models, should be employed to further optimize the feature set or compared with the method in this study. (2) The use of techniques such as SHAP (SHapley Additive exPlanations) to analyze the marginal effects and interaction effects of variables should be explored in depth into the driving mechanisms of carbon emissions. (3) This study lacks a detailed uncertainty analysis of the model's predictions. Future research could incorporate methods such as Monte Carlo simulations or bootstrap resampling to assess the uncertainty of the model's predictions and provide a more comprehensive understanding of the reliability of the results. (4) This study should consider the potential impacts of policy changes, technological advancements, or other factors. It should conduct scenario analysis to explore carbon emission trends under different policy and technological development paths. (5) For the Abstract, it should be more logical and indicate the research meaning or objective, existing problems and main contributions.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Ready for publication.

Author Response

衷心感谢您对我们工作的积极评价，您的认可是对我们莫大的鼓励。