Machine Learning for Data Center Optimizations: Feature Selection Using Shapley Additive exPlanation (SHAP)

Round 1

Reviewer 1 Report

Excellent, well researched work; evidence based - to an extreme - which is very good; theoretically sound; the data could, perhaps, have been triage-ed a bit more hierarchically - so that the reader can grasp the concept and the results more clearly; at times, it comes across as very dense in terms of data, the breadth of it and the volume of information - so the point of using AI methods - very appropriately - does not come across very clearly; this is for the future papers of the same authors

Author Response

Dear Reviewer

thanks for the comments  the revised and improved version in attachment,

best,
Marta Chinnici (on behalf of authors)

Reviewer 2 Report

First please be more clear in the sentence: " cluster with 2,0832 cores" ( in Abstract). also in Table 2, was write "excavation time". I guess is "Execution time". Please recheck the entire document. It is understandable that drafting mistakes can creep in, and in final form can be some, but we must reduce  to minimum.

The paper is interesting, but was choose only RF and XGB prediction models in analysis. These are very good methods, but are other methods such as Support Vector Machines (SVM) compatible with SHAP approximation methods? Apparently it is, but does it give the same results as RF and XGB? In order to reveal the performance of the proposed method (SHAP - which is rarely discussed in literature), it would be good if more possibilities for verifying the SHAP method analyzed.

Author Response

Dear Reviewer,

thanks for the comments, the revised and improved version in attachment,

best,
Marta Chinnici (on behalf of authors)