Application of Interpretable Machine Learning for Production Feasibility Prediction of Gold Mine Project
Round 1
Reviewer 1 Report
This paper presents an interesting application of machine learning techniques to assess the feasibility of putting gold mining projects into production. The authors develop a workflow involving data imputation, model building, and evaluation of model interpretability. The topic is relevant and the use of advanced ML methods like Miceforest and SHAP in the geology domain is novel. However, there are some aspects of the methodology and results that need further clarification.
Major comments:
The introduction provides good context about use of ML in geology, but the rationale for using Miceforest specifically for missing data could be elaborated. Why is it better suited than other imputation techniques?
More details are needed on the data pre-processing and feature engineering steps. What transformations were applied and why? Were correlated variables removed? This can impact model performance.
The results demonstrate improved accuracy with imputed dataset, but insights into model errors/limitations are lacking. Were any error analyses done? Important to assess cases where the model fails.
While SHAP values indicate feature importance, how were optimal SHAP thresholds determined to classify projects as production feasible or not? This is an important aspect for practical application.
The discussion is focused on interpreting model results. More critical analysis comparing the proposed approach to existing methods is needed. Can the workflows be improved further?
Minor issues:
Avoid excessive usage of acronyms like MICE, PMM, without writing full forms initially.
The abstract could highlight the key findings and implications better.
Enhance visual representation of results - SHAP summary plot text is unclear.
In conclusion, this is an interesting application paper, but lacks sufficient methodological details and critical discussion. Addressing the comments can improve the robustness and novelty of the work.
Minor editing of English language required
Author Response
请参阅附件。
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Dear Author(s),
The attached revisions should be made.
Best regards.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
In this manuscript, the authors applyed the Miceforest imputation algorithm to handle missing data and use a Random Forest model to predict the feasibility of moving mining projects into production. The methods employed yield commendable prediction results, and the paper is well-structured and coherently written. However, some aspects need clarification before the manuscript can be recommended for publication.
1. The authors have not specified how the dataset was partitioned. It is crucial to share this information to evaluate the model's performance correctly. In general, if the dataset is sufficiently large, I would recommend random splitting and cross-validation.
2. There are missing mathematical equations in the manuscript. For instance, the equation used to calculate the SHAP value Φi in Eq. (3) should be explicitly stated in the text.
3. Some references in the manuscript do not follow the correct citation formats. The authors should pay careful attention to the referencing style. For instance, in line 48, the authors referred to 'Mehrdad Davran (2021)', but typically, only the last name should be used when citing an individual. Additionally, if a citation is included in the bibliography, it is unnecessary to state the exact publication year in the text.
I think the English is well composed.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Thank you, all my comments are satisfied!
Minor editing of English language is needed!
Reviewer 3 Report
The authors have removed my concerns and the manuscript can be published with current form.
The English is fine.
