The Classification of Inertinite Macerals in Coal Based on the Multifractal Spectrum Method
Round 1
Reviewer 1 Report
The paper proposes an approach to analyze microscopic images with heterogeneous natural using the MF-DFA method. Three texture features have been extracted by the multifractal properties which have demonstrated to have anti-noise and anti-blur capability. A classification model with RBF-SVM classifier has been built to distinguish the 120 microscopic images of inertinite macerals in coal.
The submission addresses a very interesting research topic, which is relevant for the journal.
The idea is well described through a rigorous mathematic, and the experiments are convincing.
The paper has the potentiality to give a good contribute to the research.
Nevertheless, I suggest to improve the introduction and conclusions Sections.
The introduction should be enhanced with the benefits of using the machine learning approach in addressing numerous problems belonging to different fields. In this respect, I suggest to mention the following paper:
"Detecting unfair recommendations in trust-based pervasive environments", Information Sciences, Volume 486, June 2019, Pages 31-51. doi: 10.1016/j.ins.2019.02.015.
In the conclusions section, some detailed discussion about the impact of the results in real scenarios should be added. Moreover, future works need to be enhanced.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The Authors describe the use of automatic methods to identify macerals of the inertinite group. Because of the complex structure of the macerals, this is a difficult task to accomplish. Nevertheless, the methodology proposed by the Authors allows to achieve very good results.
An important value of the paper is the use of modern methods to identify hard coal macerals. It is worth noting that the results obtained by the Authors are similar to the best results described in other papers. This may indicate that 95% of correct recognitions are currently the optimal result.
While reading some questions and ambiguities arise. I suggest to the Authors that they consider whether to explain them in the paper.
1)
How many images were analyzed? 120 (Chapter 6 - Conclusion), 60*8 = 360 (Chapter 2) or 20*8 = 160 (Chapter 5.1)
2)
In the paper, the Authors used images with a resolution of 227 x 227 pixels - why they were so small?
3)
How were coal areas selected for imaging? Were the photos taken in such a way that there was only one maceral on them?
Quantitative analysis of coal involves identifying macerals at random places. As a result, areas where different macerals overlap are also analysed. In this case, the automatic analysis is ambiguous and difficult to implement. Have the authors tried such analysis?
4)
Did the Authors use only one type of coal for analysis? How the method will work when different coal is being analyzed.
5)
I think the results would be better presented if the axes in Fig. 4 will have the same range (e.g. alpha min: 1.6-2.0; alpha max: 1.0-4.0, fmax: 1.94-2.0).
6)
The Authors write that by observing Fig 4 "we can find" that "different macerals are separated in the space". Is this true for all macerals? (e.g. for Funginit and Secretinite?). I know that placing 8 macerals on one chart will make the chart illegible. But, I would suggest a detailed description of what separation looks like in such case. Is the separation still very good or are some classes overlapping?
Author Response
Please see the attachment
Author Response File: Author Response.pdf
