All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Thank you for systematically addressing the reviewers' comments and congratulations on the acceptance of your manuscript.
[# PeerJ Staff Note - this decision was reviewed and approved by Sebastian Ventura, a PeerJ Computer Science Section Editor covering this Section #]
The authors have now provided the key insights I requested in my previous review. A casual reader can now, I think, make sense of some important results that were glossed over.
No comment.
No comment.
I'm satisfied with the responses to my queries. I think the paper exhibits clearly, in this new version, the importance of its content.
Dear authors,
Thank you for submitting your work to PeerJ Computer Science. After completing the evaluation of your manuscript the reviewers recommend its reconsideration following major revision. I invite you to resubmit your manuscript after addressing the comments made by the reviewers. In particular:
1. Proofread your article for typos and grammatical mistakes (as suggested by both reviewers).
2. Provide a more appealing motivation and explanation for your work (as suggested by reviewer 1).
3. Extend and enhance the presentation of background concepts (as suggested by both reviewers).
4. Compare the results with a benchmark (as suggested by reviewer 1).
5. Include examples to better illustrate the scope and limitations of the proposed methodology (as suggested by reviewer 2).
6. Provide a discussion that helps to explain the observed behavior (as suggested by reviewer 2).
I hope you can complete the recommendation changes in a revision of your article.
Best,
Ana Maguitman
[# PeerJ Staff Note: The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at copyediting@peerj.com for pricing (be sure to provide your manuscript number and title) #]
The paper analyzes the impact of news on the direction of changes in stock price, exploring machine learning techniques and using data from Russian media sources. From the prediction of the price direction, the study identifies a simple trading strategy and analyzes risk-adjusted performance metrics.
- The article is well-written, and the English language is adequate. The structure and format of the paper help the reader. However, a revision of the text may improve the article (i.e. "The figure D illustrate").
- The theoretical background could be enhanced since the literature on forecasting the direction of stock price movement is abundant. Some discussion about market efficiency could make the text more appealing, linking the practical approach of the study with finance theory. The research problem implies a hypothesis that the market is not efficient (at least in a semi-strong context).
- Some references are incomplete, such as those in lines 470, 482. Therefore, a careful review of the metadata of the references is important.
- The method is adequately described. In addition, many methodological issues are addressed in the paper, enriching the discussion.
- The study discusses Granger causality and risk-adjusted performance metrics and provides results for different combinations of algorithms and data sources, making the paper more interesting.
- The trading strategy from the study is based on a simple long position on stocks that are more likely to have a price appreciation over the next week. However, this strategy could be explained better. For instance, is there a recomposition of the portfolio each week? If yes, how is it done? Does the positioning in the stock take into account closing prices? Are transaction costs computed?
- Since the Russian language is not usually addressed in papers, it might be interesting for readers to have a brief description of the similarities, differences, and challenges in using machine learning to analyze languages derived from the Cyrillic alphabet.
- The study reports results from a variety of stocks, media sources, and algorithms. It would be interesting to compare the results with a benchmark, for instance, a buy-and-hold strategy or a broad stock market index.
- Therefore, the results could be more informative and more compelling for the readers.
This paper presents a method to relate the evolution of stock quotes and relevant textual economic information, aiming to provide a tool to maximize the gains of traders. They use close to two hundred thousand news articles (from Russian sources) and 39 stock time series from 2013 to 2021. They use
Sharpe's ratio ( representing the additional amount of return of a risky asset over a risk-free one that an investor receives per unit of increase in risk) to compare their model with other 26, drawn from the literature, finding that their approach performs better than the existing alternatives. Furthermore, for more than 70% of portfolio stocks they find that the Granger correlations established by their method are high, indicating its ability to capture the causal relation between news and stock values.
The paper is well written (with only a few scattered typos and infelicities). The literature review seems adequate and the graphics both conceptual and statistical are correct and help the reader to understand the procedure and the results.
On the other hand, the paper is rather terse, with very short discussions of highly relevant concepts, ranging from assumptions to conclusions.
The methodology seems to be sound, both in the treatment of textual information as well as in the use of time series. But the presentation of the experiments carried out and of the results obtained is hastily presented. I think some examples would be useful to understand the scope (and limitations) of the methodology presented here. I would also like to see some discussion of what could explain the drastic drop in causal power of the method in 2018 that can be seen in Figure 8.
While I can see the potential of this approach, I think that more information is needed to provide a sound assessment. Figure 9 is highly relevant here and thus deserves further discussion both about its meaning and on how a trader could make a difference using this approach.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.