Evaluation of Machine Learning Models for Ozone Concentration Forecasting in the Metropolitan Valley of Mexico

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. Grid search is the core of this paper, it deserved some detailed introduction, but not even a citation about it.

2. Personally I don't like machine learning, or Grid search in this kind of topic, because no matter what kinds of data you feed to the algorithm, it will turn out to something, then you can not really relate the data to the results. For example you talk a lot about the data decide ozone at the beginning, but Grid search is blind, it just tries to find a good metric. So what I suggest is that if you can connect the hyperparameter to data, it will be better, for example, some algorithm is overfitted by Grid search, so which hyperparameter  can make it  a good fit? find it and explain it.

Comments on the Quality of English Language

The English looks good, but it is a little difficult to follow, I don't know what is the problem, maybe too many abbreviations.

Author Response

Thank you very much for your feedback and your efforts in reviewing the article. Attached is the file with responses to your comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors present a research that evaluates different models such as Random Forest (RF), Support Vector Regression (SVR), and Gradient boosting (GB) to forecast Ozone (O3) concentration 24 hours in advance, using data from monitoring networks.

Overall, the paper is well written and quite clear on the method, analysis and conclusions. The authors clearly identify their options and how they obtained the presented conclusions.

As for comments, I would like authors to characterise the monitoring dataset, in terms of missing values and outliers, and its impact on the results.

Another issue I find intriguing is that authors mention performance issues but do not specify any application requirements. Is this forecasting for operational use (near real-time)? If so, how their system would operate in terms of gathering data from monitoring network and feed into ML models?

Finally, ozone concentration is already measured by any entity using a sensor? Or is it by computational model?

 

 

Author Response

Thank you very much for your feedback and your efforts in reviewing the article. Attached is the file with responses to your comments.

Author Response File: Author Response.pdf