Cognitive Biases on the Iran Stock Exchange: Unsupervised Learning Approach to Examining Feature Bundles in Investors’ Portfolios
Round 1
Reviewer 1 Report
First of all congratulations on submitting the paper. There are some comments which could improve the paper are given below:
1) Usually if the authors of the manuscript are from the same department, it's enough just to write the department once, and to give the emails below of both authors.
2) It would be nice to see the main results (1-2 sentences) at the end of the abstract, it would be useful for the reader's first impression.
3) I would suggest removing "machine learning" from the title and all text of the manuscript, because authors I think do not understand the definition of ML at all, or at least there are some mistakes in the text which lead to my opinion. There is no mode or something where the model training has been performed, it is just the clustering algorithms that have been used to cluster the data. In some places, authors write the ML techniques k-means, etc, in other ML tools k-means, etc. It defiantly not a tool, it is an unsupervised learning method. Just better write maybe more general definition like AI methods or something, it would be less lie : )
4) There are no good arguments given about why k-means and other methods have been chosen at all. These three methods used in the "approach" could be easily changed by using just one unsupervised method self-organizing maps (SOM), it can cluster, visualize, and dimension reduction) - everything that has been made in the manuscript. For example, doi.org/10.15388/NA.16.4.14091. At least in the introduction could be analyzed various methods for clustering, for dimension reduction, to show the authors bigger scope of this field.
5) In the Introduction paragraphs formatting is different from the rest of the text.
6) From lines 49 - 65 there are a lot of references without a big sense. Just a lot of references to show that something was searching for information, but the rule is better less, but more described or analyzed.
7) At the end of the Introduction it would be nice to see the structure of the manuscript.
8) I think because of the reason that all authors' are from the Faculty of Economic Science a lot of statistics are given, that's good, but at least for me, to a person who more works on classification and clustering methods it wondering does it really enough 104 data instance for good experimental investigation and the results a well-validated? It is more discussion question. So maybe before the conclusions, it would be good to have a new section where the limits or threats of such experimental investigation have been described.
9) Maybe I missed but I think I do not find the reference to the tool which has been used for visualization, investigation, et., it would be nice to write because I think it is not the author's tool.
10) The references formatting does not meet the requirements of the journal, need to fix it.
11) Some Images have to be maybe separated because now they're so small that it is difficult well to see the results, or maybe my glasses are too weak for me. For example, the figure 6 clustering results are presented so narrowly each other, hard to see the data ID.
12) One more strange thing, the approach is based on k-means, PCA, but used ant hierarchical clustering, just to see the pairs or why at all?
My general opinion about this manuscript is, the manuscript needs some more details, more arguments in some parts of the manuscript. Now it is just really applied science when well know methods simply are used in one field. It is the right journal, but even for it needs to improve the descriptions and become more clear in the usage of the methods.
Good luck with the submission.
Author Response
Dear reviewer, thank you for your all comments. All of them were included and in our opinion really improved the quality of the paper. In the file attached we list your points with our answer and give the paper in track mode version to show all corrections we made.
All the best, Authors
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper investigates the occurrence of cognitive biases in different groups of investors on the Tehran Stock Exchange. This study looks at a few common fallacies, including confirmation bias, loss aversion, the gambler's fallacy, the availability cascade, the hot-hand fallacy, the bandwagon effect, and the Dunning-Kruger effect. Furthermore, this paper derives investor profiles that are characterised by similar fallacies and differ significantly depending on their age, stock market experience, and perception of market trends. It was discovered that inexperienced investors who are pessimistic about market development and have made significant investments are the most vulnerable group to almost all analysed biases. Old and experienced investors, on the other hand, are only affected by the Dunning-Kruger effect, are more optimistic about trends (hot-hand bias) and verify information prior to making decisions (confirmation bias), while other biases are not visible. Young-experienced investors with a large investment are generally less vulnerable to all biases and much less risk-averse.
In the title, “Cognitive biases on the Iran Stock Exchange: machine learning approach to detect feature bundles in investor's portfolio” have highlighted machine learning while the contribution of the machine learning in this paper is not strong enough and a method only has been employed. In addition, “machine learning “is a very wide area. Therefore, it is better to use specific keywords in the title. Consequently, I recommend revising the title of this paper.
The abstract of this paper is not professional at all, some findings have been presented arbitrary. In addition, the contribution of the study is not clear in the abstract. Also, please avoid using “We” in the abstract of the paper. In addition, some sentences are not clear and not professional, for example, “What is interesting, gender does not matter.”, it is like telling a story, not a scientific paper.
The introduction should completely be revised. The flow and the logic of the introduction are very unclear, and it is not clear what is the main aim and objective of this paper. In addition, gaps and the contribution of this research should be clearly highlighted in the introduction.
The literature review of this paper should be complete be polished. The authors used some factors, but it is not clear how they selected these factors. In addition, in some sections such as “Availability cascade” I can not see any reference. It seems many parts of this section are coming from the authors' knowledge, not from the literature!!!
The authors fail to present the research design in a systematically way. In addition, the reference for figure 1 has been missed.
The authors used the K-means method, what is the justification for using this method while there are many more powerful methods like ANN in the literature that can handle the nonlinear nature of the problem. In addition, validation and verification for using this method should be provided and a comprehensive test should be conducted to evaluate this method.
Author Response
Dear reviewer, thank you for your all comments. All of them were included and in our opinion really improved the quality of the paper. In the file attached we list your points with our answer and give the paper in track mode version to show all corrections we made.
All the best, Authors
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Thank you for taking suggestions into account. But there are still some mistakes, such as:
1) In the title it is enough just the "unsupervised learning", no need to write "clustering".
2) Some minor English mistakes are left.
3) In some places "-", in other "–".
The novelty of the manuscript is weak, but I think the manuscript fits in the journal of applied science. It is more the application of the existing method in one field.
Author Response
Dear reviewer, please find attached the corrected version of the paper in track-mode and answers to your points attached.
best regards, Authors
Author Response File: Author Response.pdf
Reviewer 2 Report
The quality of the figures, particularly figures 1 and 2, should be approved. The paper can be accepted after as=dressing this issue. Congrats!
Author Response
Dear reviewer, please find attached the corrected version of the paper in track-mode and answers to your points attached.
best regards, Authors
