Comparative Analysis of Clustering Algorithms and Moodle Plugin for Creation of Student Heterogeneous Groups in Online University Courses
Round 1
Reviewer 1 Report
A research project to facilitate student group work in heterogeneous groups
The key assumption is that heterogeneous groups are better than homogeneous groups, based upon one supporting reference - not convinced this is necessarily correct. As this is the crucial foundation of this research more supporting literature should be referenced to make the case for the need of the research!
Are learning outcomes improved???
What characteristics are used to determine heterogeneity? The LMS interaction data utilised is simply focused upon frequency and volume of student interaction with specific elements of the LMS - not convinced this translates to different characteristics that are useful in group formation or dynamics!
The research seems more driven by creating a Learning Analytics project rather than being fundamentally concerned with improving student learning outcomes. There is no discussion of learning theory or pedagogy to validate or situate the goal of the research.
There is no analysis in the article provided to prove that “This increases the chances of success for all students in 725 group works within the online course.”
There is no discussion of the implications of the research for ethics consent from the participating students.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
In this article, the authors propose an intelligent plugin for a Modular Object-Oriented Dynamic Learning Environment that creates heterogeneous groups using Machine Learning.
The approach taken by the authors confuses the reader to the maximum
level.
A simple application of 6 well-known clustering algorithms is made.
The article has a low mathematical background and does not provide any innovation.
The article has significant weaknesses and raises questions.
More specifically, I have to point out the following:
1)The introduction section needs enrichment. The article is missing relevant references to the approach. The review of the state-of-the-art is insufficient and not up-to-date.
2)What are the main challenges in this domain? What are the limitations of the previous works, which motivate the current study?
3)The authors apply:1.K-means 2. Mean-Shift Clustering 3. Agglomerative Clustering 4. Density-based spatial clustering of applications with noise (DBSCAN)5. Gaussian Mixture Models Clustering 6. Self-Organizing Map (SOM). What are the criteria for their selection?
The algorithms could be presented in pseudocode form, accompanied by the necessary mathematical correlations and appropriate documentation through references.
4)There is no convincing description of the dataset in which the experiments are performed. There is no information in Table 1.
What is the dataset size? What are its quality features? How is clustering done?
5)What is the implementation environment of the experiments?
6)After running the different clustering algorithms, a Silhouette analysis is made. What are the criteria for this selection?
Ιs it accompanied by the required literature and references?
7)The results section describes Moodle plugin used to generate heterogeneous groups. Is it useful to have a description of a known application in the results section?
8)It is difficult to understand how the clustering results came about.
How it turns out that K-Means and MeanSift algorithms are superior?
9)The results remain without compelling evidence of research novelty. The discussion of them is lacking.
10) Which are the limitations and the potential issues of this research? The academic implications of this study are not reported.
11)The conclusion section could be further improved to highlight the clustering results. It is not presented any numeric information and extensions of the study.
12)The use of English is low. Several syntactic, grammar and punctuation errors need correction. Also, the article has many extensive sentences. Punctuation is essential.
To sum up, this article cannot be published. It has problems with innovation, presentation and methodology.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The authors need to clarify what is the main scientific contribution of this work. If the contribution is the development of moodle plugin, The title of the paper be updated to reflect the same.
If the contribution is 'Comparative analysis of clustering algorithms for creation of heterogeneous groups', a more detailed statistical comparison needs to be done. For this, internal, external, and relative cluster validation can be attempted. A methodical comparison and justification are required as to which Silhouette method and correlation metric are chosen.
The current comparison relies exclusively on Silhouette analysis, which is a form of internal validation. The same can be strengthened by testing Dunn Index.
For external validation, the correlation has been used. However, a small number of clusters combined with a linear trend suggest spurious co-rrelation. The same needs to be tested. An alternative is to use the regression approach i.e. ANOVA. Alternatively, the corrected Rand index or Meila’s variation index VI can be explored.
For more methods, Rendón, E., Abundez, I., Arizmendi, A., & Quiroz, E. M. (2011). Internal versus external cluster validation indexes. International Journal of computers and communications, 5(1), 27-34. can be referred.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The article is now much improved and the goals and outcomes of the research more explicit.
Reviewer 2 Report
I have no additional remarks on the revised version.
The authors have addressed my concerns.
Reviewer 3 Report
The revised work represents the scientific effort as well as the results in effective manner. The comparative results of clustering are compelling. Good work.