A Real-Time Data Analysis Platform for Short-Term Water Consumption Forecasting with Machine Learning

Round 1

Reviewer 1 Report

This study presents a web-oriented platform to collect in real-time water consumption data and uses two machine learning approaches of LSTM and BPNN to forecast consumption. In general, the topic of this work is interesting and falls within the scope of the journal. The main issue with this manuscript is the lack of organization. For example, the introduction is mostly about methodology, while there is a very short and weak literature review in section 2, the reference number starts from 4, font sizes in figures are not consistent, and etc. I think this manuscript could be considered for publication after major revisions. Below, I’ve listed some comments to help the authors improve their work.

 

L5: “leads to unevenly-spaced time series” not clear

L7-8: Wired sentence, please rewrite

L 13: What do you mean by accurate precision? Please state model performance metrics.

Do not use abbreviations in keywords

L19: Why does the reference number start from 4?

L19 and 21, references should be inserted at the end of sentences

Section 1 (introduction) is in fact a part of the methodology, and thus, this manuscript does not have an introduction and does not cover the relevant literature. The authors must provide a literature review and introduce the current state-of-the-art methods and explain the novelty of their work compared to the literature.

L52- 62, These lines should be moved to the introduction.

Introduce sigma in Eq. 1

L83, please justify why a high resolution is beneficial. It seems that this work considers a time step of 1 millisecond. Is such a high time resolution really needed?

L84-85: “educes the energy consumption on the meter side” why?

L98: SQL and other abbreviations, please define the abbreviation the first time they appear in the text.

Table 1, computation time is also a function of the processor. Please state the number of cores and clock speed of the CPU used for your computations.

In Table 2, the authors have reported the consumption time in the order of 10^6 milliseconds. Please explain why didn’t you use a time step longer than 1 millisecond?

Tables 1 and 2: A relatively large number of neurons is used in each layer. As a large number of neurons may decrease the performance and increase the computation time, please explain how did you optimize the structure of your neural nets.

Fig. 5, please provide some information regarding the average number of household/population or the number of apartment units that consumed these water volumes.

Please use consistent font sizes in the figures.

L205: Typo, Conclusions

L206-2016: these lines are somewhat the repetition of your methodology. Please mainly focus on your findings

It is not very conventional to insert a period “.” At the end of an equation. I suggest removing them.

Author Response

• L5: “leads to unevenly-spaced time series” not clear

The sentence has been improved.

 

• L7-8: Wired sentence, please rewrite

The sentences have been simplified and clarified.

 

• L 13: What do you mean by accurate precision? Please state model performance metrics.

The sentence has been completed.

 

• Do not use abbreviations in keywords

Acronyms have been replaced.

 

• L19: Why does the reference number start from 4?

The reference numbers have been updated.

 

• L19 and 21, references should be inserted at the end of sentences

Done.

 

• Section 1 (introduction) is in fact a part of the methodology, and thus, this manuscript does not have an introduction and does not cover the relevant literature. The authors must provide a literature review and introduce the current state-of-the-art methods and explain the novelty of their work compared to the literature.

We have added and discussed 3 new references about the topic in the introduction.

 

• L52- 62, These lines should be moved to the introduction.

The first paragraph of the introduction has been completed to introduce a literature review.

 

• Introduce sigma in Eq. 1

Done.

 

• L83, please justify why a high resolution is beneficial. It seems that this work considers a time step of 1 millisecond. Is such a high time resolution really needed?

Explanations have been provided (because of industrial specifications).

 

• L84-85: “educes the energy consumption on the meter side” why?

Explanations have been provided.

 

• L98: SQL and other abbreviations, please define the abbreviation the first time they appear in the text.

Done.

 

• Table 1, computation time is also a function of the processor. Please state the number of cores and clock speed of the CPU used for your computations.

Details have been added at the beginning of Section “Water consumption forecasting”

 

• In Table 2, the authors have reported the consumption time in the order of 10^6 milliseconds. Please explain why didn’t you use a time step longer than 1 millisecond?

Done.

 

• Tables 1 and 2: A relatively large number of neurons is used in each layer. As a large number of neurons may decrease the performance and increase the computation time, please explain how did you optimize the structure of your neural nets.

The neural architectures and parameters have been find out by trial-and-error. This has been written done in the revised version of our article.

 

• Fig. 5, please provide some information regarding the average number of household/population or the number of apartment units that consumed these water volumes.

Additional details

 

• Please use consistent font sizes in the figures.

Figures are now clear (for example, Fig. 4 and 5 habe been largely improved).

 

• L205: Typo, Conclusions

Done.

 

• L206-2016: these lines are somewhat the repetition of your methodology. Please mainly focus on your findings

Ok, done.

 

• It is not very conventional to insert a period “.” At the end of an equation. I suggest removing them.

OK, done.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper is interesting and has some merits that can lead to publication.

Nonetheless, there are a few issues that should be fixed before.

1 The reference Secton is not updated;

2 For example, much work has been already done with LSTM and deep learning technologies in the field of water consumption (and related issues). 

3 Not only, but even with LSTM the path is not straightforward, with the need of humans in the loop, based on the literature I better know. To this aim, authors could consider for example this paper: Marco Roccetti  Giovanni Delnevo Luca Casini, Paola Salomoni: A Cautionary Tale for Machine Learning Design: why we Still Need Human-Assisted Big Data Analysis. Mob. Networks Appl. 25(3): 1075-1083 (2020).

Definitely, I would like to see a more comprehensive a well-balanced discussion on the use of deep learning in the "water" field".

Author Response

• 1 The reference Secton is not updated;

 

• 2 For example, much work has been already done with LSTM and deep learning technologies in the field of water consumption (and related issues).

 

 

• 3 Not only, but even with LSTM the path is not straightforward, with the need of humans in the loop, based on the literature I better know. To this aim, authors could consider for example this paper: Marco Roccetti Giovanni Delnevo Luca Casini, Paola Salomoni: A Cautionary Tale for Machine Learning Design: why we Still Need Human-Assisted Big Data Analysis. Mob. Networks Appl. 25(3): 1075-1083 (2020).

 

Items 1 to 3: The suggested reference has been cited. 2 other references have been added and discussed.

 

• Definitely, I would like to see a more comprehensive a well-balanced discussion on the use of deep learning in the "water" field".

The first two sections have been completed in this direction.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed my comments.

Reviewer 2 Report

The paper has greatly improved in quality