Review Reports

Round 1

Reviewer 1 Report

General: The paper is well structured and written in a plain and easy to follow manner.

I have little concerns (which doesn't mean they are wrong, just opinions to think about).

Methodology used: RNN is based sorely in "purchases". It doesn't attempt to classify customers. Although CF approaches might tend to be common (but at the same time a bit obsolete), clusterization of customers (based of number of people living at home, children, tastes, money, etc) by using algorithms like "K-means" and adopting auto adaption strategies can open a different strategy of "product recommender" that can also be very interesting (didn't find in the literature any references regarding to this).

 

My second concern is that your study used data that has already been published, but in the reserch design, you decided to exclude the top and lower 10%. How does it affect the experiment in reality? The top products are the most likely to appear in the predictions, and therefore subsecuent products might not be appearing in those predictions. And lower products might simple be products with longlife (like oil, cereals, etc) that customers buy once a while but it is important to recommend them to buy. Because your study is based in RNN, it is not possible to "estimate consumption" to recommend these products that are bought once every "many - more than 5" purchases.

 

Results are clearly explained, but I would appreciate to consider the "simplification made in the research" in the conclusions. Why did you decided to exclude those items? Did you run the RNN without excluding them? How it affected the results?

 

Overall is a good paper that could also have other implications in environment (reduction of waste), recommendations based not only in purchases but also in "similar seasonal products", etc.

Author Response

Thank you for your valuable comment.

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The abstract has to be rewrited not by including the main components from the paper structure but by describing the paper goals, objectives and findings. This abstract format, used by authors is rather a paper outline than an abstract. Other thing is related to the sustainability subject. Is not clear how this particular subject, defined in this paper, could be related to the sustainability issue. There are missing any phrases or words that could clarify this link.

I recommend to replace the title Related works with Literature Review. In terms of used approaches for the recommender system is described only the Collaborative filtering concept and then the transition is made towards the Multi-period product recommender system using RNN. There are no mentions regarding other approaches such as: Content-based filtering or Multi-criteria recommender systems. Why were no details made in this regard, since the content-based filtering concept is very well linked with the modelling of user's preferences?

Also there a lack of things that could underline the features for the fresh food market comparising with other kind of markets (for instance automotive market, electronic devices market, ITC market etc.). To be very rigorous this (i.e. fresh food market) is mentioned within the paper only in 5 cases! These clarifications can help to understand the adoption of a certain/proposed solution and the objective approach for the indicated results.

In terms of data description is mentioned a situation related to a transaction data from Fresh Delivery Service Company in USA, published in 2017 (data analysis competition platform Kaggle). Is not clear why these data are representative for the research itself! Is this enough to certify the paper findings?! (by comparison why is not considered the case of other country, other platform, other year/period and so on?!). 

A question related to the practical aspects is also the following: It can be said or explained in terms of the duration of time(minutes, hours etc.) the differences between type T time points.

Another issue is the readability of certain graphic constructions. It would even be advisable for certain graphics to be designed in a different way (for instance figure 4, 6b, 7b and 9). Also for the figure 4 are required more explanations in order to understand better the differences between the experimental results of various optimization algorithms (including the meaning of the axes used in the graph). 

Like a recap, every single observation/comment must be taken into consideration and treated accordingly (including the issue of sustainability). In the same time the results and the conclusion has to be focus on the authors contribution related to the paper objectives/research itself.

Author Response

Thank you for your valuable comment.

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I appreciate the changes made and the comments attached. For the most part of my observations, they can be considered as improved things and scientifically satisfactory treated. However, there are some questions about the usefulness of the solution and how it will be implemented. 

At present I do not see how the proposed idea (and solution) will be operable!

It is not clear to me how this solution actually will work in the real market. How the solution can be reconfigured when major changes in consumer behaviour occur? What timeframe is foreseen for an effective implementation of the solution and how can the upgrades (required) be upgraded? Which kind of steps have to be carried out in implementation for an effective operation?

By responding to the questions and offer some relevant answers can bring even more clarity regarding the need to prove the sustainabilty!That's at least at the level of future ideas, further research etc.

Author Response

I appreciate the reviewer for the valuable comments about our manuscript. Your comments regarding this paper are replied as follows.

1. I appreciate the changes made and the comments attached. For the most part of my observations, they can be considered as improved things and scientifically satisfactory treated. 

reply: Thanks for your good evaluation. 

 

2. However, there are some questions about the usefulness of the solution and how it will be implemented. At present I do not see how the proposed idea (and solution) will be operable!

reply: Thanks for your comment. And I apologize the lack of detailed explanation of the usefulness of our suggested system and operation of our system. They are explained in the following answers, so please refer the following two answers. 

 

3. It is not clear to me how this solution actually will work in the real market. How the solution can be reconfigured when major changes in consumer behaviour occur? What timeframe is foreseen for an effective implementation of the solution and how can the upgrades (required) be upgraded? Which kind of steps have to be carried out in implementation for an effective operation?

reply: Thanks for your valuable comments. According to your comment, we have explained the operation of our suggested system, and update of the system, in  final section in brief. 

They are
(... In applying the suggested recommendation methodology in the real market, it will be necessary to set the update frequency of the recommended model. In the actual stage of operation, the model is suggested to be updated daily or weekly to reflect the changing preferences of users to increase the accuracy and diversity of recommendation results. But the exact update frequency is to be decided by several experiments with real data set, and may be different from the characteristic of items, number of items, number of customers, and the average time between two sequenced purchases. ...)

 

4. By responding to the questions and offer some relevant answers can bring even more clarity regarding the need to prove the sustainability!That's at least at the level of future ideas, further research etc.

reply: Thanks for your valuable comments. According to your comment, we have explained the sustainability issues in the perspective of market manager, and in the perspective of customer, respectively, in the final section. 

They are

(... In terms of sustainability, highly accurate multi-point recommendations that reflect the changing customer preferences can help market managers prevent products with very short shelf life, such as fresh vegetables, from being discarded. Furthermore, as our suggested multi-point recommender system is a kind of decision support system, it could help customers to make rapid routine decisions and save their time and money. Moreover, we expect that our study contributes the customers’ reduction of food wastes by inducing planned consumption. ...)