Review Reports
- Sergio Hernández-Casas1,
- Luis Felipe Beltrán-Morales1,* and
- Victor Gerardo Vargas-López2
- et al.
Reviewer 1: Xiaoying Zhang Reviewer 2: Anonymous Reviewer 3: Christopher Chun Ki Chan Reviewer 4: Nguyen Quoc Khanh Le
Round 1
Reviewer 1 Report
Review manuscript titled “Price forecast for Mexican Red Spiny Lobster (Panulirus spp.) Using Artificial Neural Networks (ANNs)” by Hernández-Casas et al.
This manuscript introduced the application of Artificial Neural Networks in the price forecasting of Mexican red lobster and proved its predictive power compared to the ARIMAX model. The manuscript provided some valuable results which are helpful for fishery resource management.
However, I have a few comments that the authors should clarify.
The average monthly export price of the Mexican lobster is kind of confused to me. Normally, the price should not be a continuous varied function from 0 to a top price, then drop to 0 as shown in figure 1 in time. For example, there would be a low limitation on price like whatever 10 or 12, but definitely not from 0. Because your data is continuous, 0.001(any non 0 number on your plot) dlls/kg is not a real lobster price.
Thus when you set the price data as your modeling target, the results which the model has predicted is not that convinced in the results part. Even compared to the price in figure 1, it's hard to tell your plot of the “Forecast” in figure 6, and the NARX predicted value is way too high compared to the marketing data, A model that predict a real price of 30 as 48, is not acceptable in the market. It is not helpful but maybe misleading the fisher management in practice. A research that how to get a good result of NARX has been discussed in the paper “Juan Zhang et al., 2019 in JOH (https://doi.org/10.1016/j.jhydrol.2019.123948)”. NARX is a powerful tool in ANN, which should give you better fitting results. Please refer that paper for details information. In addition, for the purpose of comparison, I suggest adding the results of ARIMX in the figure.
Some detailed comments:
In page 2, line 84-85, please specify what “other prediction model” specifically refers to. It is also worth noting that you introduced the application of ARIMA in fisheries resources price forecasting in the introduction, but then used ARIMAX model for comparison. Please be consistent.
In page 3, Table 1, it looks like a bracket is missing.
Please increase the resolution of the picture 2-3.
In page 4, line 155-159, I don't think a test accuracy of more than 70% is a very high accuracy. Maybe it's because there are too few training samples.
In page 8, line 251, there is no fig 7 in the manuscript, please clarrify.
In page 9, figure 6, it seems that the blue line is real price according to the information in Figure 1. Please check. Additionally, because this figure covers the information of Figure 1, I suggest deleting Figure 1.
In page 7, table 2, what are the correlation coefficients of other variables? Please further explain why 0.41 is used as the basis for division.
According to the correlation in table 2, it can be seen that VHK has a higher correlation coefficient of 0.86 than PMex and PAus, but the NARX model with one variable has a lower R2. Please further analyze the reasons.
Reviewer 2 Report
Dear Authors,
the subject of the article is very interesting, however, to ensure good quality of the publication, I suggest making some improvements.
- Please describe in detail the functions used in the neural network. Among other things, describe and explain the selection of the number of hidden neurons, activation functions, and interneural weights.An example of a description of a neural network can be found in numerous publications. E.g.:https://doi.org/10.3390/app112110414https://doi.org/10.3390/app10051897https://doi.org/10.1016/j.geoderma.2019.06.016
- Please explain why in the diagram Fig. 6 the price of the lobster over a period of about 35 months is not correctly predicted.
- Can the prices be adequate for lobsters from other regions of the world as well? Can the neural network tool proposed by the authors also be used to predict lobster prices from other regions of the world? Will the input and output parameters be the same?
- Please improve the bibliography section by supplementing it with a wider range of publications presenting the prediction task using artificial neural networks.
- Please add more observations on the novelty of your research.
Reviewer 3 Report
The main idea of the paper:
1. Originality and significance of the research
Using ANNs to predict the price of fishing resource to support better utilization of the resource. Essential for small-scale fisheries.
2. Technical and theoretical correctness
ANNs applied to a relatively narrow field within a specific region. ANNs are applied correctly.
3. Readability of the paper
Introduction has different fonts.
Too quick of an attempt at background information, please be more detailed, provide insights, and not just clump up 5-6 references together. Expand a bit and elaborate, but provide relevant and useful insights to the reader.
Figures need to aligned to the margins better.
Headings need to be differentiated, different font? numbering? eg. "Autoregressive models"
Goodness of fit, reword, many headings need rewording and rethinking, use commonly used words for this type of work please, and better alignment of your numbering system.
Mean squared error formula is not exactly a crucial formula to warrant entry in this document.
Elaborate more on the data used, what affect does it have on the ANN
4. Evaluation result
Better spacing of results figures. Provide indicators or lines on the actual chart to better help describe what is intended.
5. Scope of the work.
Ok, within scope.
Reviewer 4 Report
Overall, the manuscript is poorly prepared and written. There are a lot of errors that can be found, for example, no affiliation or author information, no declaration section, the fonts are inconsistent among paragraphs, and many grammatical errors or typos, ...
For those reasons, even not yet considering the scientific quality of the study, I could not recommend it further for publication.