Machine Learning Methods to Estimate Productivity of Harvesters: Mechanized Timber Harvesting in Brazil

Round 1

Reviewer 1 Report

Abstract - Abstract is not clear at all and must be re-written to adequately communicate the core summary of the research.

Key words - Hardly any of your key words are in the Abstract.

Introduction - The introduction is not providing an adequate background of the research problem. It goes straight into data analysis questions and uses terminologies of which which the reader does not yet understand the applicability. For example - data wrangling is described - but there is no indication of what this is and why it is necessary. A full background must first be given of what data wrangling is even necessary. There are many different types of forest machines - most are self-propelled, so this general description is inadequate. If you focussed on cut-to-length, then why not say "mechanised cut-to-length" machines, or if the focus in only on harvesters, then why not "timber harvesters"?

General - The article seems to have been put through a translator - resulting in English that does not flow and does not always make sense. I stopped commenting on the language structure as this article needs to go through an English language editor.

Materials and methods:

Not described with sufficient clarity and depth of understanding to allow the reader to have confidence in the research that was carried out.

Dataset: Line 96 - the researchers dive straight into instances, but still no proper breakdown is provided of what data is actually being collected. Mention is made of categorical, numerical and ordinal - but what does this apply to. Be absolutely clear about what was measured, any categories, frequency of measurement etc. For example, it says that machine availability is measured, but no information is provided on this?

Line 104: What was the experience of the operator? Was the operator experiences on all the different machines because a learning operator will give variation in results.

Data wrangling - still not adequately described regarding exactly what was done in the data wrangling process and what criteria were used to wrngle. More detail needed on SMOTE.

Machine learning - not enough detail to give the reader/reviewer sufficient comfort that the tests/methods selected are appropriate. For example 123 - "We tested machine learning algorithms and selected them according to their performance in predicting the productivity of forest machines" - no additional information is given. We just have to assume that this is all in order. Where did the machine learning algorithms come from? Why did you select what you did?

Results: The language use again makes it difficult to concentrate on the science being explored. The results are very difficult to read with a poor explanation of what is being done and why. The flow of the information is poor and the reader is left behind as to why certain results are being presented.

Discussion - Difficult language use makes engagement with the arguments difficult. This needs to be addressed before the science can be examined again properly. Because the initial sections do not adequately explain the various methods used and why, it now becomes difficult to interpret the results.

Conclusion - It is stated that productivity can be predicted. Some figures are provided to this effect. But this seems to be the results of statistical processes and there does not appear to be adequate testing of the soundness of the predictions made. If this is just visual then it needs to be stated, otherwise the predictions need to be described through modelling and testing of the fit of the models.

 

 

Comments for author File: Comments.pdf

Author Response

Dear reviewer, 

Please see the attachment. 
Kind regards,

Author Response File: Author Response.docx

Reviewer 2 Report

Based on data sets affecting logging productivity, such as the individual tree volume, terrain slope, and availability of hours of use of the machine, the productivity of self-propelled forest machines of the harvester type is predicted by using a variety of machine learning algorithms, which assist forest manager in making decision through strategic, tactical and operational plans. The study result is of reference value for forestry managers to realize the intelligent management of mechanized harvesting of wood. However, the paper has the following problems:

1. The data sets used in this study are large in scale, but they are not scientifically and rigorously indicated and screened from the source of data and the specific processing method of data in the paper. It only gives descriptive explanations, which should be supplemented.

2. Although a variety of machine learning methods are applied to predict the productivity of forest machines, the existing algorithms are not be analyzed, selected or improved by authors in the applicability of the productivity of forest machines in the paper. In addition, compared with single algorithm such as Extra Tree regressor or CatBoost regressor, blending ensemble algorithm and stacking ensemble algorithm is no significant improvement in terms of evaluation indexes such as determination coefficient and mean absolute percent error, and the calculation amount and time are increased. Therefore, the conclusion is of little significance.

3. It is difficult to associate Fig.3 and its description with productivity indicators, please supplement.

4.  This paper is hard to read and understand, and there are some grammatical errors,  need to be improved with the assistance of English editing.

Author Response

Dear reviewer, 

Please see the attachment. 
Kind regards,

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Since the authors answered all the questions raised by the reviewer, I suggest that the manuscript need some minor revisions.

Specific modifications are as fellows  according the manuscript titled "forests-1770368-revision without track change" :

1. Delete "metrics" in 21 line, 

2. "cofficient of determintation" will be  changed to "determination cofficient"  in 22, 24 lines;

3. Change "your" into "our" in line 27;

4.  Change "experimentation and maximizing the model's performance" into" maximizing the model's performance by conducting experiment" in line 27;

5. Change "the inference" into "to infer" in line 264.

 

Author Response

Dear reviewer,

Please see the attachment.

Kind regards,

Author Response File: Author Response.pdf