A Garlic-Price-Prediction Approach Based on Combined LSTM and GARCH-Family Model

Round 1

Reviewer 1 Report

In this paper, authors proposed a hybrid model using LSTM and GARCH models for prediction of garlic price. My overall assessment is this paper is a well-written article. The proposed model is  interesting to some readers  in time series prediction using neural networks.

 

Minor comments

 

1. Aurthors need to review more references related to the hybrid model based on GARCH models and neural networks.

I recommend that authors cite the following works.

 

Aras, Serkan. "Stacking hybrid GARCH models for forecasting Bitcoin volatility." Expert Systems with Applications 174 (2021): 114747.

Koo, Eunho, and Geonwoo Kim. "A Hybrid Prediction Model Integrating GARCH Models with a Distribution Manipulation Strategy Based on LSTM Networks for Stock Market Volatility." IEEE Access 10 (2022): 34743-34754.

Huang, Yumeng, Xingyu Dai, Qunwei Wang, and Dequn Zhou. "A hybrid model for carbon price forecasting using GARCH and long short-term memory network." Applied Energy 285 (2021): 116485.

Seo, Monghwan, and Geonwoo Kim. "Hybrid forecasting models based on the neural networks for the volatility of bitcoin." Applied Sciences 10.14 (2020): 4768.

Kakade, Kshitij, Aswini Kumar Mishra, Kshitish Ghate, and Shivang Gupta. "Forecasting Commodity Market Returns Volatility: A Hybrid Ensemble Learning GARCH‐LSTM based Approach." Intelligent Systems in Accounting, Finance and Management 29, no. 2 (2022): 103-117.

Zolfaghari, M. and Gholami, S., 2021. A hybrid approach of adaptive wavelet transform, long short-term memory and ARIMA-GARCH family models for the stock index prediction. Expert Systems with Applications, 182, p.115149.

 

2. I think that it is hard to see Table 5.  Please, modify the Table 5. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The current study reflects an interesting idea. However, I have a few issues which convinced me to recommend a revision.

1.      Motivation, challenges, and contribution are clear but the comparative analysis confuses me. According to the author “The experimental results show that the prediction performance of the combined LSTM and GARCH family models containing volatility characteristic information of garlic price is generally better than that of the single model. The combined LSTM model incorporating GARCH and PGARCH models (LSTM-GP) has the best performance for predicting garlic price in terms of evaluation indexes, such as mean absolute error, root mean square error, and mean absolute percentage error. The combined model of LSTM-GARCH family has better results in garlic price prediction and can provide support for garlic price prediction”. Here a question rises why you did not compare your work with the existing studies? There are many algorithms both machine learning and deep learning but why the authors selected LSTM and GARCH.

2.      I will highly recommend shifting at least 2 paragraphs from the introduction to the related work section, so the introduction will be more clear for future readers.

3.      Please cite the particular dataset instead of putting the link here “The experimental data source of this paper is from the garlic industry chain big data platform (http://www.garlicbigdata.cn/) developed by the Big Data Center of Shandong Agricultural University.”

4.      Figure 1 caption is not clear. Please briefly explain.

5.      Equation no. 6 represents?

 

6.      Table 5 represents the model accuracy and comparative analysis. I will highly recommend briefly explaining each of the accuracies in paragraph separately. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors present a garlic price prediction method based on LSTM.

In my opinion the paper is well-written. 

Anyway, the paper has some serious flaws.

1. What is the novelty? It seems just another price prediction method based on well-known machine learning model (i.e., LSTM). The authors should emphasize the gaps in the literature and their novelties in the introduction. Furthermore, the contributions to the vast price prediction literature should be emphasized.

2. The related work section is weak. It's just a listing. Authors should compare point to point with every work. Moreover, I suggest the authors to change the structure: first, they should discuss about price prediction in general (many of the previous works can be blindly applied here), then focus on price prediction in agriculture and lastly specifying the combined models. A table summarizing main features of every work against the current one could be helpful to grasp primary information, similarities and differences.

3. The authors state that they use a combined model. They should expand the section and ground to existing literature on combined models (not only those for agriculture and price prediction). I suggest authors to look at these papers (but there are others): https://doi.org/10.1007/s00521-022-07454-4, https://doi.org/10.1109/BIBM.2015.7359925.

4. The benchmark models are not enough to justify publication. The proposal of the model should be motivated by a careful comparison with existing methods for price prediction, at least those available and reproducible, and in the scope of the paper (which should be clarified more). For example, how does this proposal compare against https://doi.org/10.1007/s00521-022-07543-4 ?

5. The authors should add an expected impact section at the end of the paper.

6. Where is the code?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Authors have replied to all my comments.

The paper has improved a lot since the last version.

Now, the contributions is clearer.

Still, the expected impact section should be improved in my opinion:

in this work the authors focused on price prediction, but how can this be used in real life? For example, in https://doi.org/10.1007/s00521-022-07543-4, the authors employed a broad set of intelligent agents that not only predict the prices but actually trades.
This could be discussed as future works / expected impact of the project.

Furthermore, the combination of models has not been related to multi-modal learning, multi-view learning and integration in the machine learning literature.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf