Optimizing Automated Trading Systems with Deep Reinforcement Learning

Round 1

Reviewer 1 Report

The lack of presentation of the findings of the study in a concise manner in the paragraph that comes right before the last one in section 1. While this is a very wonderful thing, it is necessary to identify the contributions that have been made. It would be helpful if you included those contributions to the list. Unfortunately, that was not the case with you.

I would want to propose the establishment of a new SECTION 2 that would be solely devoted to the reporting of earlier empirical research studies (Related Work section) and the establishing of the gaps related with such studies. To be more specific, I will propose the inclusion of a summary table of the findings from the previous research.

It is recommended that the present sections 2, 3, and 4 be consolidated into a single section known as section 3: Research Methodology and then rewritten so that the current section 2 can be the conceptualising portion (also known as the "Trading Strategy"). In addition, as part of the primary objective of the study, you are going to need to provide a justification for why it is necessary to optimise automated trading systems. Because you only have one citation related with that portion of the study, even though this should be one of the most significant aspects of this research. In a similar vein, in order to effectively draw the concept of your current section under "Parameter Optimization," there should ostensibly be a robust justification of the concepts. You just provide one citation, and you don't draw anything fresh in connection with the idea you're discussing.

Because you need to provide your freshly formulated notion in the exact same place where the current sections 3.1, 3.2, and 3.3 are located, the current sections need to be reorganised. The "Algorithm 1 Parameter optimization with DRL" section of the paper has the most significant flaw in the research. You will need to provide a line-by-line explanation of the "Algorithm" in order to fulfil this requirement.

The description of Figure 1 in your current section 4, which is titled "The interaction between the trader and the market," is not entirely clear. In Figure 1, the arrows and the bullet points do not move in synchrony with one another. Labeling is required from the very beginning all the way through to the very conclusion.

The information in Section 4.1 is jumbled and difficult to understand. To begin, there is neither a detailed description nor any evidence regarding how you obtain your dataset.

In addition, I will propose the following two subsections: (specifically, the sections titled "DATASET" and "EXPERIMENTAL PROCEDURE") In the part labelled "Dataset," you are required to demonstrate the data you utilised by providing a screenshot or photograph, followed by an explanation of the captions. In the section on the experimental technique, you will need to provide specifics regarding the experimental circumstances (scenarios) and the stages that were carried out (screenshot or picture). Then provide an explanation of the computational resources that were utilised.

Simply taking into account the experiment described in reference [3] is not sufficient justification for you to use the "ADAM algorithm for weight optimization." This is due to the fact that you need to first link your concept to the requirement for optimization, then present as many optimization algorithms as you can, and finally justify your decision to use the "ADAM algorithm for weight optimization."

When you split or partition your dataset for "Training" and "Testing," you need to have a solid reason connected with the ratios that you employ for your segmentation.

The current section 4 is the part of the paper that is least strong. Why, when you're labelling and numbering your equations, do you overlook the equation that begins with your current section 2, "trading Strategies," and then proceed to start just in the current "Section 3.1?" In addition, despite the fact that you formulated and presented all of the equations in sections 2 and 3, why didn't you indicate how they were implemented in the "result section"? They are of no help to you or your research if this is the case.

The representation of the results also does not point out where the measure of the errors of optimization can be drawn. The work does not set out a metric to ensure that the optimization measurement would or could be associated with any optimization error

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The reviewed paper presents interesting research concerning an automated trading system and the algorithmic trading strategy. The proposed title of the paper represents the content of the manuscript. The principal results, the major conclusions, and the research methodology are briefly described in the abstract. The abbreviations used within the manuscript are listed at the end of the paper. The authors briefly presented the scientific background and the state-of-art of the investigated issues in the introduction section and referenced the literature. However, they are encouraged to refer to more up-to-date papers and indicate the novelty of their research in comparison to the ones presented in the literature review. The applied Relative Strength Index is described but it is a well-known and commonly applied indicator. On the other hand, multiple techniques to optimize the parameters of a trading strategy using the Relative Strength Index are successfully applied. The results are adequately discussed and well presented. The conclusions are correctly drawn based on the obtained results.

Author Response

The authors appreciate valuable comments of the reviewer. In the revised version, a Related Work section is added with more up-to-date studies. The novelty and urgency of the study are also presented and compared with previous studies. The article has been restructured to better understand the research problem. The algorithms are explained in detail. The experiments are presented simply and without loss of generality, the results can be applied to other technical indicators.

Round 2

Reviewer 1 Report

I am not fully satisfied with the responses, this indicates that you need to look over the full corrections again, as almost all of the responses, with the exception of the "Algorithm," include errors.

The research contributions are not included in the "Introduction section" in the format of a list of bullet points. You have to do this.

Author Response

• The research contributions are included in the "Section 1. Introduction".
• Studies on objective functions and evaluation metrics used for financial data are added to "Section 2. Related work".
• "Section 3.4. The trading system" is rewritten with clearer explanations and discussion of the reasons for choosing the values and algorithms.

Reviewer 2 Report

The authors have addressed all the issues that I indicated in the first-round review. Therefore, I recommend accepting the paper.

Author Response

• The research contributions are included in the "Section 1. Introduction".
• Studies on objective functions and evaluation metrics used for financial data are added to "Section 2. Related work".
• "Section 3.4. The trading system" is rewritten with clearer explanations and discussion of the reasons for choosing the values and algorithms.