Machine Learning Fusion Multi-Source Data Features for Classification Prediction of Lunar Surface Geological Units
Round 1
Reviewer 1 Report
This paper develop a method of combining machine learning algorithms with features of data to build a classification model for classification prediction of geological units. I think this is a very interesting experiment that has intrigued planetary science researchers.
I have a question that I am not clear about. How many sample labels are there and what improvements have been made to the trained model?
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The paper is very well written, the results are fine and provide a significant progress for the lunar research community.
A very important question remains (for me). MAybe this is not an issue here, however I wonder about the fusion method for the individualsoruces. Obviously, the different sources have different spatial (and temporal) resolutions. How is this accounted form? The authors describe some gridding (to Chang'e-2 high-resolution image data [line 100]), and this seems fine, however:
> What are the resolutions (spatial and temporal) for the individual data sources?
> Is it safe to postulate 18492 and 23326 sample points, respectively [line 106]. Are these sample points significant in a statistical manner or is this not an issue here?
There should be some mechanism to co-locate the data points and *also* account for the resolution in space (and time). Maybe this is not an issue here, however if o, it is not well explained.
----------------------------------------------------------
Layout:
Fig. 6: Too small, please cut into separate figures to have each figure larger. Also please increase font size.
----------------------------------------------------------
Typos:
"pyrthon" -> python (?) (line 78)
"Confusion Matrix" -> confusion matrix (line 156ff)
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Review of Remote Sensing manuscript 1944515
This is a well-written paper that will be of interest to the planetary remote sensing community. The results of the study are well-written and presented. My comments are generally minor in nature.
1. Line 78: did you mean to say “Python”?
2. Table 1. In 2 places you have written “DOM”. Did you mean to write “DEM”?
3. Line 137: add “the” after “build”
4. Line 142: delete “of”
5. Lines 137 and 138: I am not sure what you mean 2,6, and 8, and 2 and 4 in these lines. Should these numbers be associated with some kind of unit?
6. Discussion for Figure 7: while you correctly note that XGBoot performs best, you might also want to note that some of the other techniques perform nearly as well, with average accuracies within 1% of the XGBoot results, so XGBoot is the best, but not by very much. This may encourage the use of other techniques which would be nearly as good as XGBoot.
7. Line 361: do not capitalize “Extensibility”
Author Response
Please see the attachment.
Author Response File: Author Response.docx
