Unsteady Multi-Element Time Series Analysis and Prediction Based on Spatial-Temporal Attention and Error Forecast Fusion

Round 1

Reviewer 1 Report

In this manuscript, the authors propose a new DA-RNN method to predict CHl-a in advance in order to analyse the Harmful algal blooms (HABs) issue. Fourteen independent variables have been chosen in this study. Several other methods (such as LSTM, Seq2Seq and BP) have been selected to use as comparisons in this study. 

However, about this manuscript, I have several concerns, detailed as below:

(1) The influence of each independent variable has not been investigated in detail here. I wonder how much mutual information they have to each other.

(2) Generally, the performance (RMSE and MAE) should be decreasing with the increasing of the prediction intervals for RNN methods. However, in Table 2 & 3, 12-hour-intervals has the best RMSE and 18-hour-intervals has the best MAE. The authors are suggested to do some investigations about it.

(3) The manuscript is not logically organized. The demonstration way for figures and tables is lack of description. Their results are also lack of discussion at times. More analyses are required (both statistical and physical).
. The authors are suggested to improve it.

Based on those above concerns, I think this manuscript may not be suitable for publication in future internet before a major revision.

Author Response

Dear Reviewer,

Thank you for your comments concerning our manuscript entitled "Unsteady Multi-element Time Series Analysis and Prediction Based on Spatial-Temporal Attention and Error Forecast Fusion" (Manuscript ID:futureinternet-690795). Those comments are valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researchs. We have stuided comments carefully and have made correction which we hope meet with approval. Revised portion are marked in red in the paper. The main corrections in the paper and the responds to your comments are uploaded in the attachment.

Please see the attachment.

Thank you and best regards.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper makes affirmations unsupported by a scientifically justified justification, but only numerically comparative, on particular results obtained for the compared methods. The conclusions are not unequivocally drawn and this part must be improved.

Author Response

Dear Reviewer,

Thank you for your comments concerning our manuscript entitled "Unsteady Multi-element Time Series Analysis and Prediction Based on Spatial-Temporal Attention and Error Forecast Fusion" (Manuscript ID:futureinternet-690795). Those comments are valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researchs. We have stuided comments carefully and have made correction which we hope meet with approval. Revised portion are marked in red in the paper. The main corrections in the paper and the responds to your comments are uploaded in the attachment.

Please see the attachment.

Thank you and best regards.

Author Response File: Author Response.pdf

Reviewer 3 Report

I notice that there are many paragraphs and phrases similar to literature; although some parts there is a citation. However, I suggest a complete review in this regard; Define acronyms once in the text (the first time it is cited); The definition of some acronyms is missing, e.g., LSTM, ANN, ML, RNN (Recurrent Neural Networks), DA, etc.; There is a need for a complete grammar revision: "Fussy BP method"? etc.; Use, whenever possible, citation of benchmark publications, e.g., BP [A]; Standardize the use of the (normal or only italic) style of variables/parameters; Vectors are not represented in the usual form of mathematical notation; Citations are missing in important parts of the text, e.g. in some equations, fuzzy sets, artificial, neural networks, etc.; I suggest to the Authors to highlight (objectively) the innovation of this proposal concerning literature.

Reference

[A]  WERBOS, P. J.  “Beyond regression: New tools for prediction and analysis in the behavioral sciences”, Ph.D. Thesis - Harvard University, 1974.

Author Response

Dear Reviewer,

Thank you for your comments concerning our manuscript entitled "Unsteady Multi-element Time Series Analysis and Prediction Based on Spatial-Temporal Attention and Error Forecast Fusion" (Manuscript ID:futureinternet-690795). Those comments are valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researchs. We have stuided comments carefully and have made correction which we hope meet with approval. Revised portion are marked in red in the paper. The main corrections in the paper and the responds to your comments are uploaded in the attachment.

Please see the attachment.

Thank you and best regards.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors addressed most questions I had in the first round. Therefore, I recommend this paper to be published in the present form.

Reviewer 2 Report

The changes made did not lead to a significant improvement of the work.

Reviewer 3 Report

-