Analysis of SAP Log Data Based on Network Community Decomposition
Round 1
Reviewer 1 Report
Dear authors:
The manuscript present relevant issues.
Still, some minor reviews should be performed.
In general, the paper looks more practical and practice-oriented. Additionally, the design of the arguments and the way of communication throughout the manuscript may cause a distraction for a reader. Therefore, it is essential for the paper to rearrange the manuscript by including more arguments, justifications of the executed tasks and discussions as well as taking them just to the main point to be emphasized.
Overall, the article presents different sections unbalanced according to the number of pages, so some sections should be amplified such as introduction and conclusions; and the methodology section should be reduced. In this regard, it would be desirable to increase the conclusions section, since the quality of the work done is worth it.
As for the Abstract, it is so generic, better to make it more original and consistent to the paper I would make the Abstract sharper: what is the problem, what have you done, what is the contribution.
Regarding the next points, the first paragraph of the introduction may be more appropriate to be introduced into the methodology section. Moreover, the second paragraph should contain more extended and clear the objectives.
Precisely, the methodology first paragraph shows the objectives more clearly and it may be more advisable to be into the introduction. Likewise, I would like to know what it is the data source. Please also, justify why LRNet algorithm is used instead of other algorithms. As for figure 1, the letter should be amplified, due to the fact that it is hard to read the texts. In addition, according to the construction of network and clusters identification more numerical results of the obtained cluster should be shown so as to check if the results obtained are right or wrong. Besides, please take into account the possible typography mistakes or missing references (i.e. Error! Reference source not found-line 195). Please consider the possibility to delete “Result” in line 196 since this word is shown in bold type and at the beginning of a paragraph. Indeed, there is a specific section so-called “Results”. It is compulsory to set a number for every figure in the text, line 204 “…is shown in Figure…”. This is a common mistake made in the text, please correct it. Again it is shown the word “Result”-line 211, it would be applied the same comment.
Regarding the section of results, again Table 5 shows a text “Figure” several times but the figure number is not shown and on line 294 also “Figure” without number, after that “Error! Reference source not found” is displayed again. As for Figures 4 and 5, it is impossible to correctly read the text (numbers) in the circles. In this sense, it would be advisable to amplify the size of the letter for the legend in Figures 6, 7, 8 and 9 and other figures.
Generally speaking, the figures are shown with the wrong number, the enumeration is not correlative. In addition, it would be advisable to show more numerical results in order to check the research validity, even though some of them are shown in the Appendix A. Else, it is really difficult to put the finger on the matter.
Regarding the discussion section, I would like to know what the limitations are and what future research should be done in the future.
As for conclusions I said before it should be larger.
In this regard, another point to take account in a general way is to please give more explanations about the findings and so as to support your arguments, add more citations (there are only 8 citations) and of course, please articulate what the main contribution of this paper is. Furthermore, please also mention the limitations and future research directions.
Lastly, as for English language and style, English language and style are fine/minor spell are required.
Kind regards,
Comments for author File: 
Comments.pdf
Author Response
Dear reviewer, the point-by-point response is attached.
Author Response File: Author Response.pdf
Reviewer 2 Report
lines12-17 need to be proof read, the structure is not a correct use of english
l31, laic is too obscure, please rephrase
l38, please briefly describe SAP log data first
l44, one sentence does not suffice for a paragraph
l64, the sentence started there is too succinct to make sense to the reader. It must be expanded upon in a manner which emphasizes the actions or data model taking place
Fig1, the Object Attribute table is not described in enough detail by this point. The reconstruction from those entries is not clear. It is conceptual but not clear.
The object attribute networks should relate to the work of:
@article{bothorel2015clustering,
title={Clustering attributed graphs: models, measures and methods},
author={Bothorel, C{\'e}cile and Cruz, Juan David and Magnani, Matteo and Micenkova, Barbora},
journal={Network Science},
volume={3},
number={3},
pages={408--444},
year={2015},
publisher={Cambridge University Press}
}
@article{newman2016structure,
title={Structure and inference in annotated networks},
author={Newman, Mark EJ and Clauset, Aaron},
journal={Nature communications},
volume={7},
pages={11863},
year={2016},
publisher={Nature Publishing Group}
}
As well, how is the work related to:
@article{basole2018ecoxight,
title={ecoxight: Discovery, Exploration, and Analysis of Business Ecosystems Using Interactive Visualization},
author={Basole, Rahul C and Srinivasan, Arjun and Park, Hyunwoo and Patel, Shiv},
journal={ACM Transactions on Management Information Systems (TMIS)},
volume={9},
number={2},
pages={6},
year={2018},
publisher={ACM}
}
@inproceedings{dong2017metapath2vec,
title={metapath2vec: Scalable representation learning for heterogeneous networks},
author={Dong, Yuxiao and Chawla, Nitesh V and Swami, Ananthram},
booktitle={Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining},
pages={135--144},
year={2017},
organization={ACM}
}
@article{van2014multivariate,
title={Multivariate network exploration and presentation: From detail to overview via selections and aggregations},
author={Van den Elzen, Stef and Van Wijk, Jarke J},
journal={IEEE Transactions on Visualization and Computer Graphics},
volume={20},
number={12},
pages={2310--2319},
year={2014},
publisher={IEEE}
}
@article{liu2014ploceus,
title={Ploceus: Modeling, visualizing, and analyzing tabular data as networks},
author={Liu, Zhicheng and Navathe, Shamkant B and Stasko, John T},
journal={Information Visualization},
volume={13},
number={1},
pages={59--89},
year={2014},
publisher={Sage Publications Sage UK: London, England}
}
l84, what models and methods from 'data mining'?
l91, very interesting, but how do you get all the data into numerical format? are there not 'factors'/'levels'/categorical variables in the data? How do you handle that?
l107, 'waves' is there not a better choice of terminology?
l125-128 need to be in an algorithmic format
l135, of a patterns model
section 2.3, IQR is not the standard outlier detection approach for research, more novel exploration is required
section 2.4, This is too basic and how are the network outliers being computed? Surely not with univariate IQR?
l177, what is the 'business meaning'?
eq1, should the first 2 terms not precede the third?
eq2, how is t_rj computed?
Fig 6-10, the x-axis is too small to read
Why do the figures after 10, start from 1 again?
This is interesting work and a different approach but it needs a significant amount of proofreading and clarity improvement. Also a more extensive literature review is required.
Author Response
Dear reviewer, the point-by-point response is attached.
Author Response File: Author Response.pdf
Reviewer 3 Report
This is an interesting paper. In the paper, the authors proposed a new method to analyse the running process and participants’ behavior from the process log by integrating data mining and machine learning and network analysis and visualization techniques. They used their new approach to construct the complex network and detect the patterns in the network using SAP log data. The results show the data visualization, network and community detection techniques are important for business decision support systems.
Major issue:
In introduction section, the authors should give more background information, clearly stated the research objective, and defined their research problem and included some work has been done by others. (or related works)
It will be appreciated if authors could give more deep discussions about their finding and its implications and limitation in Discussion section.
Materials and Method section contains lots of information. The structure of this section needs to be improved. Some parts are repeated in the different subsections. There are two 2.4 sections and two 2.5.1.1 sections but each has different content.
It is better to refine the Results section. This section can be changed to Experiments and Result. All the information about Data collection and description, experiment set up, results, and analysis could be included in this section.
Minor issue:
The paper’s readability is OK but there are typos and grammar errors that need to be fixed and some sentences are too long and hard to be understood. It would be appreciated if the authors could do some proofreading before the final submission.
The captions[SZ1] of Table 1 and Table 2 should be above the tables.
Please check page 16 and 17. Figure 1 should be Figure 11 and Figure 2 should be 12.
[SZ1]
Author Response
Dear reviewer, the point-by-point response is attached.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The improvements are satisfactory although more work could be done in terms of customizing the output for the target format of the journal publication rather than copying the default output into this article. That would have helped but it is not absolutely necessary.
The references now touch on the minimum requirements and although it would have been beneficial to the reader to see how this work fits into the larger image, it is still satisfactory.
Reviewer 3 Report
The authors have addressed my comments. The quality of the new version has improved a lot.
