Prediction Model for the Chemical Futures Price Using Improved Genetic Algorithm Based Long Short-Term Memory
Round 1
Reviewer 1 Report
The manuscript can be accepted after major revisions. I have attached my detailed comment.
Comments for author File: 
Comments.pdf
Author Response
Comments: I have carefully reviewed the paper and found that the results of the paper look interesting. Overall, the paper and subject are related to the journal's scope and interesting topic for the readers. Therefore, the manuscript can be accepted for publication after the major revisions.
- The acronyms in the title did not give a proper indication of the presented study. Therefore, it must be revised.
Response: We revised the title as per your suggestion. Thank you.
- In abstract, specially the first sentence is quite long and is hard to follow. Re-right the abstract by giving a complete indication of the problem, methodology and objectives of the work. Some numerical results must be highlighted in the abstract.
Response: We revised the Abstract as requested.
- Number of acronyms are used in the abstract without definitions. Like what is LIDPE and PP? Such mistakes must be revised in the updated manuscript.
Response: We revised them. We just use the mane of the evaluation measures as they are well-known.
- Very limited literature has been provided in the introduction. The Introduction should be revised and must consist of five paragraphs answering the following five questions:
What is the problem?
Why is it interesting and important?
Why is it hard?
Why hasn't it been solved before? (Or, what's wrong with previously proposed solutions or methodologies?)
What are the key components of my approach and results?
As the above-mentioned questions should be replied to, it will be better to write more relation to the benefits & disadvantages of the blending techniques and investigate about limitations of previous studies. Also, this part needs more explanations to state clearly the objectives & hypothesis of this study at the end of the Introduction part.
Response: We revised the introduction section to be clearer for the readers and presents the main requested information.
- Also, extend the literature, and the references section must be improved by highlighting the relevant studies such as "--" Optimal caching scheme in D2D networks with multiple robot helpers "--" TBSM: A traffic burst-sensitive model for short-term prediction under special events "--"A Few Shot Classification Methods Based on Multiscale Relational Networks "--" A Deep Fusion Matching Network Semantic Reasoning Model"--"Knowledge base graph embedding module design for Visual question answering model "--" Financial deregulation and operational risks of energy enterprise: The shock of liberalization of bank lending rate in China "--" Impact of COVID-19 on stock price crash risk: Evidence from Chinese energy firms "--"
Response: We added the suggested things.
- A graphical view of the LSTM must be provided to see how the layers are connected.
Response: We added a new figure to show the LSTM.
- Commas and fullstops must be strictly followed in the equations.
Response: Done.
- Surprisingly, no references has been provided in the explanation of basic genetic algorithm.
Response: We added the related references there.
- Revise Figure 1. It is not clear in terms of quality.
Response: We try our best to improve it.
- Table 1 to 7 has been presented without any explanation.
Response: We added more explanation regarding the given figures
- Section 4.3 must be named as results and discussion. A detailed explanation should be presented to validate the result of the designed technique.
Response: We added a new section called 4.3 to show the results are requested.
- A table should be given to explain all the acronyms and parameters/ symbols used in this paper.
Response: Added
- Captions to the figures and tables must be revised properly in the updated manuscript.
Response: Revised
- Figure 2, there are no labels and legends.
Response: Revised.
Author Response File: Author Response.pdf
Reviewer 2 Report
- The results for the model put forward in this paper, i.e., the IGA-LSTM, and the standard SVM in Table 8 are very similar. Therefore, the superiority of the algorithm proposed, if any, is limited, and that should be highlighted.
- To properly evaluate the results, a valuable statistic akin to the R squared would be helpful: e.g., the variance of the forecast error over the variance of the variable under study. Halving the 'forecast error' has a very different meaning if the original fit is already very good, i.e., over 95%.
- Instead of splitting the available sample in two sets, one for training the algorithm, the other for testing, it would be much better to enlarge the sample; since the paper deals with market prices, much longer series than the period considered in the paper must be available.
- Somehow the authors should ensure the replicability of their results -e.g., making the data set and software programs available in the supplementary material
- The question of possible explanatory variables is not mentioned; this is one of the most serious drawbacks of the AI algorithms applied to economic time-series, since they are closely interrelated: e.g., past values of related variable will surely help forecasting the variable under study.
- The non-stationarity of the series analysed is commented in line123, pg.4. The implications of that property are not immediately clear but, in any case, the authors should report the results of appropriate statistical tests to support that assertion.
- There should be a description of the paper and its sections at the end of section 1.
- The English is poor and should be thoroughly revised: e.g., 'forecastand' line36 pg. 1. is not an English word.
- Most acronyms are not defined, particularly 'LSTM' which is precisely the main technique implemented!
- Limitations of the study and possible improvements in the methodology, beyond 'application to other time series' should be pointed out more clearly and in detail.
- The title of section 4.3. is 'CONCLUSION', which is the same as section 5. Correct it.
Author Response
- The results for the model put forward in this paper, i.e., the IGA-LSTM, and the standard SVM in Table 8 are very similar. Therefore, the superiority of the algorithm proposed, if any, is limited, and that should be highlighted.
Response: We pointed out that the superiority of the algorithm proposed is limited sometimes in line 268-269.
- To properly evaluate the results, a valuable statistic akin to the R squared would be helpful: e.g., the variance of the forecast error over the variance of the variable under study. Halving the 'forecast error' has a very different meaning if the original fit is already very good, i.e., over 95%.
Response: We revised and showed the R squared and some explanation in line 230-232.
- Instead of splitting the available sample in two sets, one for training the algorithm, the other for testing, it would be much better to enlarge the sample; since the paper deals with market prices, much longer series than the period considered in the paper must be available.
Response: We've used all the sample data we can get. Linear low density polyethylene and polypropylene were officially listed on Dalian Commodity Exchange on July 31, 2007 and February 28, 2014, respectively. However, thank you very much for your advice. Next, we will obtain more samples in price prediction of other fields and use IGA-LSTM to test whether the superiority of the algorithm is related to the number of samples.
- Somehow the authors should ensure the replicability of their results -e.g., making the data set and software programs available in the supplementary material- The question of possible explanatory variables is not mentioned; this is one of the most serious drawbacks of the AI algorithms applied to economic time-series, since they are closely interrelated: e.g., past values of related variable will surely help forecasting the variable under study.
Response: The supply and demand data of chemical products are mainly weekly and monthly data, and the data lag seriously. The frequency of possible explanatory variables cannot be matched. The problem is even worse from a practical standpoint. We added the suggested things in line 285-289 to explain it.
- The non-stationarity of the series analysed is commented in line123, pg.4. The implications of that property are not immediately clear but, in any case, the authors should report the results of appropriate statistical tests to support that assertion.
Response: We added the suggested things in line 215-217.
- There should be a description of the paper and its sections at the end of section 1.
Response: We added the description there.
- The English is poor and should be thoroughly revised: e.g., 'forecastand' line36 pg. 1. is not an English word.
Response: We revised them.
- Most acronyms are not defined, particularly 'LSTM' which is precisely the main technique implemented!
Response: We revised them.
- Limitations of the study and possible improvements in the methodology, beyond 'application to other time series' should be pointed out more clearly and in detail.
Response: We added the suggested things in line 282-289.
- The title of section 4.3. is 'CONCLUSION', which is the same as section 5. Correct it
Response: Revised.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have done their level best to answer my comments. Therefore, the manuscript can be accepted for publication.
Reviewer 2 Report
the corrections implemented are sufficient in my opinion, and the paper looks significantly improved.
