Forecasting Fine Particulate Matter Concentrations by In-Depth Learning Model According to Random Forest and Bilateral Long- and Short-Term Memory Neural Networks
Round 1
Reviewer 1 Report
This manuscript is interesting, however, it needs a rigorous revision to address reviewer's comments and improve the scientific representation as outlined on the attachment.
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment
Author Response File: 
Author Response.doc
Reviewer 2 Report
This paper presents some innovations for forecasting PM2.5 in China. I believe the idea is well presented and discussed especially in the Introduction. But I am not sure if this document is considered a review or a scientific paper. Moreover, I have here some minors remarks to the authors in order to improv the paper before the publication, Please not that the following remark applied to the all sections of the papers. Here just some examples:
Minors:
Line 2: please rewrite the title and add the study area name!!
Line 6: Please define "PM2.5" by "atmospheric particulate matter (PM) that have a diameter of less than 2.5 micrometers"
Line 9: Please replace “~3 h” by “3 hours”
Line 13 please replace "0~12 h PM2.5" by "PM2.5/0~12 h "
Line 17: Please replace "PM2.5" by "PM2.5" and remove the “h” near to 1.
Line 18: keep jut the RF and BiLSTM as keywords without abbreviation meaning and add more significant keywords for this study.
Line 25: please check the exact meaning of the "PM2.5"!
Please use one form "PM2.5" or "PM2.5" in the hall document!!
Line 38: please define "PM10"!
Line 52: please define “RNN”!!
Line 62: Please put “(ARIMA)” after word “average” and add after its word “model”!
Line 67: Please the parameters “SO2, VOCs and NOx”!!
Fig1: Please add to the figure different sign to the 4 selected area with identification in the legend, identify the name of the country at the map, add the orientation, remove the word “Legend” and improve the quality of the map.
Line 118: define “O3, and CO”!
Fig2: improve the quality of the figure!!
From line 168 to 176 the description is not supported by the figure 3. Can rewrite in order to describe very well the figure 3. Especially, you are using colours that will help for the description
Lines: 190, 191 and 192: please label the equations!!
Fig4: improve the quality of the figure!
Author Response
Please see the attachment
Author Response File: Author Response.doc
Reviewer 3 Report
The article is important for the prediction of PM2.5 concentration and appropriate to the journal theme. The results are very important however, some revisions are suggested in the attached manuscript. Also should be highlighted the article importance, for human health due to the necessity of portion of early warning system, in conclusions.
Comments for author File: Comments.pdf
Author Response
Please see the attachment
Author Response File: Author Response.doc
Reviewer 4 Report
The manuscript titled "Forecasting PM2.5 concentrations by in-depth learning model according to random forest and bilateral long- and short-term memory neural networks” was investigated the prediction of PM2.5 concentration using coupling the Random Forest (RF) variable selection and bi-directional long short term memory (BiLSTM) neural net. The study showed that suggested model can be offer high accuracy of PM2.5 concentrations from 1h to 12h.
The ms fully falls within the editorial purposes of the journal and is focused on a very current issue, albeit extensively investigated in the literature. The ms is well structured and contains a large amount of very interesting and convincing data but an adequate review appears necessary before publication. I list some comments and suggestions for improvements below:
1- Line 6-10: It is preferable to suffice with one sentence to define the problem in “Abstract”.
2- The “Abstract” section should contain the parts of the manuscript that explain the problem, work methodology and results, in short
3- The keywords must not be repeated in the title.
4- The objective are not clear and need to be significantly improved.
5- It is preferable to summarize the meteorological data used through the statistical description (i.e., mean, maximum, minimum, standard deviation, skewness coefficient, and kurtosis coefficient).
6- The SVM and Tree used in the study lack explanation.
7- The splitting of data for the training and testing phases mentioned in Figure 3 is inconsistent with that mentioned in line 179-180. In general, the percentages should be as follows: 70% for training, 30 for testing.
8- I suggest make a last phase of generalization or external validation using other data not used in the phases of training and testing.
9- The amount of results reported is impressive but their critical lacks references.
10- The results of this study should be compared with previous studies.
11- Avoid to used “we” or “will” throughout the text.
12- The conclusion is not clear. What are the main findings? No need to rewrite the results in conclusion section.
Author Response
Please see the attachment
Author Response File: Author Response.doc
Round 2
Reviewer 1 Report
Authors did not address all reviewers comments, so this manuscript does not meet standards required for a scientific report.
The discussion is too brief to justify the novelty of findings, so it does not worth of publication.
Authors were asked to justify their statistical methods used in comparison with the previous commonly used procedures, however, they ignored this comment.
As a result of lacking acknowledgement of previous findings, the current findings are not scientifically reliable and thus, this paper is unsuitable for publication.
Author Response
As the academic editor suggestion, we shouldn't answer the question of this reviewer.
Reviewer 4 Report
In my point of view in this research, the researcher has done a good job. Still the authors did not address some of the comments mentioned in the previous review. Some comments:
1- Line 7-11: It is preferable to suffice with one sentence to define the problem in “Abstract”. The keywords must not be repeated in the title.
2- The objective are not clear and need to be significantly improved.
3- It is preferable to summarize the meteorological data used through the statistical description.
4- The SVM and Tree used in the study lack explanation.
5- Avoid to used “we” or “will” throughout the text.
6- The conclusion is not clear. What are the main findings? No need to rewrite the results in conclusion section.
Author Response
Please see the attachment
Author Response File: Author Response.doc
