Globalizing Food Items Based on Ingredient Consumption
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis study employs LSTM to predict the performance of food items using historical time series data. The topic is quite intriguing.
I cannot identify any innovations in the paper, as LSTM and hyperparameter optimization techniques have been extensively employed across various domains.
The experiments seem incomplete. Many state-of-the-art prediction approaches have been proposed in recent years, which the authors should have considered.
The evaluation metrics for the proposed model are also insufficient.
Comments on the Quality of English Languagegood
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors1. This paper lacks innovation.
2. The experimental comparison and results are not convincing enough.
3. How to optimize the LSTM by GA, and the parameters of GA should be presented.
4. The logic of the paper is not clear.
Comments on the Quality of English LanguageThe quality of English language is good.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsKindly look the attachment
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThe paper is devoted to the problem of food sales forecasting. The authors constructed the model based on LSTM networks and genetic algorithm and tested it on the Kaggle dataset.
The paper has serious disadvantages.
1) The abstract should indicate that the data was taken from Kaggle. In its current form, it seems that the data was collected by the authors.
2) Authors claim that “The primary goal of this research is to develop an innovative AI-powered forecasting tool that combines Genetic Algorithms (GA) with Long Short-Term Memory (LSTM) networks.” But this combination is not new. The aim of research should contain some kind of scientific novelty.
3) There is almost no literature review in the paper. It is necessary to indicate previous studies in the field of food consumption forecasting. The research gap is not clear.
In addition, descriptions of the methods used, such as LSTM, gradient boosting and genetic algorithms, should be given.
4) Please, inform the reader about the structure of the manuscript at the end of Introduction section.
5) Conceptual model of the proposed approach will be useful.
6) Subsection 2.1 - Data description is incomplete. What features were used? What was the type of features? What range? How many missing values? What was the target column? This information must be provided.
7) Subsection 2.1.1 contains only general descriptions. There is no description of specific preprocessing procedures for the data used. The same for subsequent sections: there are no specific parameters of algorithms given, no details about k-fold CV (what was k?). The parameters of the system on which the experiments were performed and the time they took are not indicated. Thus, it is impossible to reproduce the experiments conducted by the authors, which is a prerequisite for a scientific article.
What was the train/test ratio for the LSTM model? If 136 lines were used for training, then this is extremely small for a neural network. If gradient boosting was used to turn the test set into a training set, then the value of the obtained estimates is questionable regardless of further application of genetic algorithms.
8) Equations (11)-(15) must be mentioned in the text before their appearing.
9) Figures 1-2 quality is low.
10) The results are presented only as the total MSE and MAE values. It would be useful to present other results, including the variation of averaged values, as well as other metrics, for example F1, R2 and AUC ROC.
11) The reference list contains the remains of template. The same for Lines 342-344, 347-354.
12) The purpose of Appendix A is unclear. And it is not clear what the authors mean by “ingredients”. The generally accepted concept does not coincide with data from https://www.kaggle.com/competitions/restaurant-revenue-prediction
There are typos and inaccuracies in the text.
Thus, the article is written in extremely insufficient detail, which does not allow reproducing the experiment carried out in the article. The article cannot be published in its current form.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have improved the quality of this paper and solved all questions.
Comments on the Quality of English LanguageThe quality of english language is good.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsComment 7 in my review version1, was misread. Still you can compare your results with state of the art literature already published to prove efficacy of your work.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for Authors I have read the updated manuscript and have only one comment - the text in figures 2 and 5 is too small. In my opinion, the article can be accepted.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf