Next Article in Journal
Recent Advances in MXene-Based Nanocomposites for Wastewater Purification and Water Treatment: A Review
Previous Article in Journal
Drivers of Spatial and Temporal Dynamics in Water Turbidity of China Yangtze River Basin
 
 
Article
Peer-Review Record

A Stacking Ensemble Model of Various Machine Learning Models for Daily Runoff Forecasting

Water 2023, 15(7), 1265; https://doi.org/10.3390/w15071265
by Mingshen Lu 1,2, Qinyao Hou 1,2, Shujing Qin 1,2, Lihao Zhou 1,2, Dong Hua 3, Xiaoxia Wang 1,4 and Lei Cheng 1,2,*
Reviewer 1:
Reviewer 2: Anonymous
Water 2023, 15(7), 1265; https://doi.org/10.3390/w15071265
Submission received: 2 March 2023 / Revised: 19 March 2023 / Accepted: 20 March 2023 / Published: 23 March 2023

Round 1

Reviewer 1 Report

Lu et al. assess several approaches that use machine learning for daily runoff forecasting and propose an ensemble learning model based on the attention mechanism.  Overall, this is an interesting and well-written paper.  My comments below are minor with the most significant one being that the authors use R2 to denote correlation when the lowercase r should be used (R2 is the coefficient of determination).  Given there are different correlations for parametric vs. non-parametric datasets, the authors also need to clarify that these are Pearson correlations.   

Minor Comments:

1)  Line 29:  Correlation coefficient variable is r (not R2) and in various other locations as noted below.  The authors also need to mention that these are Pearson correlations.

2)  Line 73:  ANN and SVR are not defined.

3)  Line 114:  Correlation coefficient variable is r (not R2).

4)  Figure 2:  Rt, It, It-1, It-2, Pt are not defined.

5)  Line 232:  Correlation coefficient variable is r (not R2).

6)  Equation 6:  Correlation coefficient variable is r (not R2).

7)  Equation for NSE is numbered 6 but should be 7.

8)  Line 296:  Correlation coefficient variable is r (not R2).

9)  Line 301:  Correlation coefficient variable is r (not R2).

10)  Line 309:  Correlation coefficient variable is r (not R2).

11)  Figure 6d:  Correlation coefficient variable is r (not R2).

12)  Table 2:  Correlation coefficient variable is r (not R2).

13)  Line 325:  Correlation coefficient variable is r (not R2).

14)  Line 333:  Correlation coefficient variable is r (not R2).

15)  Table 3:  Correlation coefficient variable is r (not R2).

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

In this paper, a novel ensemble learning model based on attention mechanism is proposed for the daily runoff prediction problem. The proposed model has a two-layer structure with a base model and a meta model. The topic and the results obtained are relevant

I include some comments and suggestions below:

1. Abstract: rephrase and reduce it.

2. Why the authors choose Machine Learning Methods based on Random Forest, Adaptive Boosting and Extreme Gradient Boosting? No other ML methods. justify.

3. Proposed Stacking Ensemble Learning: revise it again to include the pseudo code of the proposed method. Figure 2. is not clear. 

4. Hyper-Parameter Optimization: how do you determine the values of hyper-parameters of the three machine learning models? justify.

5. Model Performance Evaluation: why do you include RMSE and MAE? one is enough. justify.

6. Results: you cannot say that the method is better than other in statistical analysis (Average and Standard Deviation). It is better to use Tests of Significance to make sure about the proposed method which is the best method. In addition, include some related works for comparison. 

Despite these shortcomings, this paper is considered to be accepted with minor revision.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Back to TopTop